Exercises for Introductory Statistics Explained

Edition 1.09

©2015 Jeremy Balka

All rights reserved
Balka ISE 1.09 2
Contents

1 Introduction 1

2 Gathering Data 3
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.2 Populations and Samples, Parameters and Statistics . . . . . . . . . . . . 3
2.3 Types of Sampling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.3.1 Simple Random Sampling . . . . . . . . . . . . . . . . . . . . . . . 4
2.3.2 Other Types of Random Sampling . . . . . . . . . . . . . . . . . . 4
2.4 Experiments and Observational Studies . . . . . . . . . . . . . . . . . . . 5
2.5 Chapter Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

3 Descriptive Statistics 15
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
3.2 Plots for Qualitative and Quantitative Variables . . . . . . . . . . . . . . . 15
3.2.1 Plots for Qualitative Variables . . . . . . . . . . . . . . . . . . . . 15
3.2.2 Plots for Quantitative Variables . . . . . . . . . . . . . . . . . . . . 16
3.3 Numerical Measures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
3.3.1 Summation Notation . . . . . . . . . . . . . . . . . . . . . . . . . . 16
3.3.2 Measures of Central Tendency . . . . . . . . . . . . . . . . . . . . . 16
3.3.3 Measures of Variability . . . . . . . . . . . . . . . . . . . . . . . . . 17
3.3.4 Measures of Relative Standing . . . . . . . . . . . . . . . . . . . . 20
3.4 Boxplots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
3.5 Linear Transformations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
3.6 Chapter Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
3.6.1 Basic Calculations . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
3.6.2 Concepts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
3.6.3 Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29

4 Probability 35
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
4.2 Basics of Probability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
4.2.1 Sample Spaces and Sample Points . . . . . . . . . . . . . . . . . . 35

i
Balka ISE 1.09 CONTENTS ii

4.3 Rules of Probability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36

4.3.1 The Intersection of Events . . . . . . . . . . . . . . . . . . . . . . . 36
4.3.2 Mutually Exclusive Events . . . . . . . . . . . . . . . . . . . . . . . 36
4.3.3 The Union of Events . . . . . . . . . . . . . . . . . . . . . . . . . . 36
4.3.4 Complementary Events . . . . . . . . . . . . . . . . . . . . . . . . 36
4.4 Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
4.5 Bayes’ Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
4.6 Counting rules: Permutations and Combinations . . . . . . . . . . . . . . 40
4.7 Probability and the Long Run . . . . . . . . . . . . . . . . . . . . . . . . . 41
4.8 Chapter Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
4.8.1 Basic Calculations . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
4.8.2 Concepts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
4.8.3 Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47

5 Discrete Random Variables and Discrete Probability Distributions 55

5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
5.2 Discrete and Continuous Random Variables . . . . . . . . . . . . . . . . . 55
5.3 Discrete Probability Distributions . . . . . . . . . . . . . . . . . . . . . . . 56
5.3.1 The Expectation and Variance of Discrete Random Variables . . . 56
5.4 The Bernoulli Distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
5.5 The Binomial Distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
5.5.1 Binomial or Not? . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
5.5.2 A Binomial Example with Probability Calculations . . . . . . . . . 60
5.6 The Hypergeometric Distribution . . . . . . . . . . . . . . . . . . . . . . . 61
5.7 The Poisson Distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
5.7.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
5.7.2 The Relationship Between the Poisson and Binomial Distributions 62
5.7.3 Poisson or Not? More Discussion on When a Random Variable has
a Poisson distribution . . . . . . . . . . . . . . . . . . . . . . . . . 63
5.8 The Geometric Distribution . . . . . . . . . . . . . . . . . . . . . . . . . . 63
5.9 The Negative Binomial Distribution . . . . . . . . . . . . . . . . . . . . . 65
5.10 The Multinomial Distribution . . . . . . . . . . . . . . . . . . . . . . . . . 65
5.11 Chapter Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
5.11.1 Basic Calculations . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
5.11.2 Concepts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
5.11.3 Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
5.11.4 Extra Practice Questions . . . . . . . . . . . . . . . . . . . . . . . 75

6 Continuous Random Variables and Continuous Probability Distribu-

tions 81
6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
6.2 Properties of Continuous Probability Distributions . . . . . . . . . . . . . 81
6.2.1 An Example Involving Integration . . . . . . . . . . . . . . . . . . 82
Balka ISE 1.09 CONTENTS iii

6.3 The Continuous Uniform Distribution . . . . . . . . . . . . . . . . . . . . 83

6.4 The Normal Distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
6.4.1 Finding Areas Under the Standard Normal Curve . . . . . . . . . . 84
6.4.2 Standardizing Normally Distributed Random Variables . . . . . . . 84
6.5 Normal Quantile-Quantile Plots . . . . . . . . . . . . . . . . . . . . . . . . 86
6.5.1 Examples of Normal QQ Plots . . . . . . . . . . . . . . . . . . . . 86
6.6 Other Important Continuous Probability Distributions . . . . . . . . . . . 87
6.6.1 The χ2 Distribution . . . . . . . . . . . . . . . . . . . . . . . . . . 87
6.6.2 The t Distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
6.6.3 The F Distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
6.7 Chapter Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
6.7.1 Basic Calculations . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
6.7.2 Concepts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
6.7.3 Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
6.7.4 Extra Practice Questions . . . . . . . . . . . . . . . . . . . . . . . 93

7 Sampling Distributions 95
7.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
7.2 The Sampling Distribution of the Sample Mean . . . . . . . . . . . . . . . 96
7.3 The Central Limit Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . 97
7.4 Some Terminology Regarding Sampling Distributions . . . . . . . . . . . . 98
7.4.1 Standard Errors . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
7.4.2 Unbiased Estimators . . . . . . . . . . . . . . . . . . . . . . . . . . 98
7.5 Chapter Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
7.5.1 Basic Calculations . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
7.5.2 Concepts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
7.5.3 Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
7.5.4 Extra Practice Questions . . . . . . . . . . . . . . . . . . . . . . . 103

8 Confdence Intervals 105

8.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
8.2 Interval Estimation of µ when σ is Known . . . . . . . . . . . . . . . . . . 105
8.2.1 Interpretation of the Interval . . . . . . . . . . . . . . . . . . . . . 106
8.2.2 What Factors A˙ect the Margin of Error? . . . . . . . . . . . . . . 107
8.2.3 An Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
8.3 Confdence Intervals for µ When σ is Unknown . . . . . . . . . . . . . . . 107
8.3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
8.3.2 Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
8.3.3 Assumptions of the One-Sample t Procedures . . . . . . . . . . . . 109
8.4 Determining the Minimum Sample Size n . . . . . . . . . . . . . . . . . . 109
8.5 Chapter Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
8.5.1 Basic Calculations . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
8.5.2 Concepts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
Balka ISE 1.09 CONTENTS iv

8.5.3 Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112

8.5.4 Extra Practice Questions . . . . . . . . . . . . . . . . . . . . . . . 117

9 Hypothesis Testing 123

9.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
9.2 The Logic of Hypothesis Testing . . . . . . . . . . . . . . . . . . . . . . . 124
9.3 Hypothesis Tests for µ when σ is Known . . . . . . . . . . . . . . . . . . . 124
9.3.1 Constructing Appropriate Hypotheses . . . . . . . . . . . . . . . . 124
9.3.2 The Test Statistic . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
9.3.3 The Rejection Region Approach to Hypothesis Testing . . . . . . . 124
9.3.4 P-values . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
9.4 Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
9.5 Interpreting the p-value . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
9.5.1 The Distribution of the p-value When H0 is True . . . . . . . . . . 128
9.5.2 The Distribution of the p-value When H0 is False . . . . . . . . . . 128
9.6 Type I Errors, Type II Errors, and the Power of a Test . . . . . . . . . . . 128
9.6.1 Calculating Power and the Probability of a Type II Error . . . . . 129
9.6.2 What Factors A˙ect the Power of the Test? . . . . . . . . . . . . . 130
9.7 One-Sided Test or Two-Sided Test? . . . . . . . . . . . . . . . . . . . . . . 130
9.8 Statistical Signifcance and Practical Signifcance . . . . . . . . . . . . . . 131
9.9 The Relationship Between Hypothesis Tests and Confdence Intervals . . . 131
9.10 Hypothesis Tests for µ When σ is Unknown . . . . . . . . . . . . . . . . . 131
9.10.1 Examples of Hypothesis Tests Using the t Statistic . . . . . . . . . 132
9.11 More on Assumptions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132
9.12 Chapter Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
9.12.1 Basic Calculations . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
9.12.2 Concepts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
9.12.3 Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142
9.12.4 Extra Practice Questions . . . . . . . . . . . . . . . . . . . . . . . 147

10 Inference for Two Means 153

10.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153
10.2 The Sampling Distribution of the Di˙erence in Sample Means . . . . . . . 153
10.3 Hypothesis Tests and Confdence Intervals for Two Independent Samples
(When σ1 and σ2 are known) . . . . . . . . . . . . . . . . . . . . . . . . . 154
10.4 Hypothesis Tests and Confdence Intervals for µ1 − µ2 (When σ1 and σ2
are unknown) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154
10.4.1 Pooled-Variance t Tests and Confdence Intervals . . . . . . . . . . 154
10.4.2 Welch (Unpooled Variance) t Tests and Confdence Intervals . . . . 155
10.4.3 Guidelines for Choosing the Appropriate Two-Sample t Procedure 157
10.5 Paired-Di˙erence Procedures . . . . . . . . . . . . . . . . . . . . . . . . . 157
10.5.1 Paired-Di˙erence t Tests and Confdence Intervals . . . . . . . . . 157
10.6 Investigating the Normality Assumption . . . . . . . . . . . . . . . . . . . 160
Balka ISE 1.09 CONTENTS v

10.7 Chapter Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 160

10.7.1 Basic Calculations . . . . . . . . . . . . . . . . . . . . . . . . . . . 160
10.7.2 Concepts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 160
10.7.3 Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165
10.7.4 Extra Practice Questions . . . . . . . . . . . . . . . . . . . . . . . 171

11 Inference for Proportions 181

11.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181
11.2 The Sampling Distribution of p̂ . . . . . . . . . . . . . . . . . . . . . . . . 181
11.3 Confdence Intervals and Hypothesis Tests for the Population Proportion p 181
11.4 Determining the Minimum Sample Size n . . . . . . . . . . . . . . . . . . 182
11.5 Inference Procedures for the Di˙erence Between Two Population Proportions183
11.5.1 The Sampling Distribution of p̂1 − p̂2 . . . . . . . . . . . . . . . . . 183
11.5.2 Confdence Intervals and Hypothesis Tests for p1 − p2 . . . . . . . 183
11.6 Chapter Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 184
11.6.1 Basic Calculations . . . . . . . . . . . . . . . . . . . . . . . . . . . 184
11.6.2 Concepts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 184
11.6.3 Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 186
11.6.4 Extra Practice Questions . . . . . . . . . . . . . . . . . . . . . . . 190

12 Inference for Variances 193

13 χ2 Tests for Count Data 195

13.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195
13.2 χ2 Tests for One-Way Tables . . . . . . . . . . . . . . . . . . . . . . . . . 195
13.2.1 The χ2 Test Statistic . . . . . . . . . . . . . . . . . . . . . . . . . . 195
13.2.2 Testing Goodness-of-Fit for Specifc Parametric Distributions . . . 196
13.3 χ2 Tests for Two-Way Tables . . . . . . . . . . . . . . . . . . . . . . . . . 197
13.4 Chapter Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 199
13.4.1 Basic Calculations . . . . . . . . . . . . . . . . . . . . . . . . . . . 199
13.4.2 Concepts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 199
13.4.3 Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201
13.4.4 Extra Practice Questions . . . . . . . . . . . . . . . . . . . . . . . 208

14 One-Way ANOVA 211

14.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211
14.2 One-Way ANOVA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211
14.3 Carrying out the One-Way ANOVA . . . . . . . . . . . . . . . . . . . . . 211
14.3.1 The Formulas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211
14.3.2 An Example with Full Calculations . . . . . . . . . . . . . . . . . . 213
14.4 What Should We Do After a One-Way ANOVA? . . . . . . . . . . . . . . 214
14.4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 214
14.4.2 Fisher’s LSD Method . . . . . . . . . . . . . . . . . . . . . . . . . . 214
Balka ISE 1.09 CONTENTS vi

14.4.3 The Bonferroni Correction . . . . . . . . . . . . . . . . . . . . . . . 215

14.4.4 The Tukey Procedure . . . . . . . . . . . . . . . . . . . . . . . . . 216
14.5 Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 216
14.6 A Few More Points . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 216
14.7 Chapter Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217
14.7.1 Basic Calculations . . . . . . . . . . . . . . . . . . . . . . . . . . . 217
14.7.2 Concepts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217
14.7.3 Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 219
14.7.4 Extra Practice Questions . . . . . . . . . . . . . . . . . . . . . . . 224

15 Introduction to Simple Linear Regression 227

15.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 227
15.2 The Linear Regression Model . . . . . . . . . . . . . . . . . . . . . . . . . 227
15.3 The Least Squares Regression Model . . . . . . . . . . . . . . . . . . . . . 228
15.4 Statistical Inference in Simple Linear Regression . . . . . . . . . . . . . . 229
15.4.1 Model Assumptions . . . . . . . . . . . . . . . . . . . . . . . . . . 229
15.4.2 Statistical Inference for the Parameter β1 . . . . . . . . . . . . . . 230
15.5 Checking Model Assumptions with Residual Plots . . . . . . . . . . . . . . 230
15.6 The Correlation Coeÿcient and the Coeÿcient of Determination . . . . . 232
15.7 Estimation and Prediction Using the Fitted Line . . . . . . . . . . . . . . 233
15.8 Transformations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 234
15.9 A Complete Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 234
15.10Outliers and Infuential Points . . . . . . . . . . . . . . . . . . . . . . . . . 234
15.11Some Cautions about Regression and Correlation . . . . . . . . . . . . . . 235
15.11.1 Always Plot Your Data . . . . . . . . . . . . . . . . . . . . . . . . 235
15.11.2 Avoid Extrapolating . . . . . . . . . . . . . . . . . . . . . . . . . . 235
15.11.3 Correlation Does Not Imply Causation . . . . . . . . . . . . . . . . 236
15.12A Brief Multiple Regression Example . . . . . . . . . . . . . . . . . . . . . 236
15.13Chapter Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 236
15.13.1 Basic Calculations . . . . . . . . . . . . . . . . . . . . . . . . . . . 236
15.13.2 Concepts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 238
15.13.3 Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 241
Chapter 1

Introduction

(No exercises.)

1
Balka ISE 1.09 2
Chapter 2

Gathering Data

J.B.’s strongly suggested exercises: 1, 2, 3, 4, 5, 6, 8, 9, 12, 15, 17, 18, 21, 24, 29, 31, 32

2.1 Introduction

2.2 Populations and Samples, Parameters and Statistics

1. Succinctly state the di˙erence between a parameter and a statistic.

2. Researchers are interested in estimating the average weight of three-year-old walleye

in a lake. A netting program yielded 52 three-year-old walleye, with an average
weight of 273 grams.

(a) What is the population of interest?

(b) What is the sample?
(c) Is the value 273 a parameter or a statistic?
(d) What is the parameter of interest in this scenario?
(e) Is it possible to know the value of the parameter of interest in this scenario?

3
Balka ISE 1.09 2.3. TYPES OF SAMPLING 4

2.3 Types of Sampling

2.3.1 Simple Random Sampling

3. An importer of used medical equipment wants to estimate the proportion of defec-

tive pacemakers in a shipment of 200 pacemakers. They open the shipping con-
tainer, remove the top 8 pacemakers, and fnd that 0% of of these 8 pacemakers are
defective. Unknown to the importer, 10% of the pacemakers in the entire shipment
are defective.

(a) What is the population of interest?

(b) What is the sample?
(c) What value given above is a statistic?
(d) What value given above is a parameter ?
(e) Would this sample above constitute a simple random sample?

4. Suppose a professor wishes to estimate the proportion of female students in a class

of 50 students. They could fnd out the true proportion without too much diÿculty,
but they decide instead to draw a sample of 10 students. Each student’s ID number
is put on a marble. The 50 marbles are put into a box, and the box is shaken
vigorously. Suppose after the vigorous shaking that the marbles can be considered to
be randomly distributed in the box. Without looking at the marbles, the professor
reaches into this box and pulls out 10 marbles without replacement. The 10 students
with these ID numbers are a sample from the population of 50 students. Which, if
any, of the following statements are true?

(a) The sample can be considered to be a simple random sample from the popu-
lation.
(b) Each possible sample of size 10 has the same chance of being selected.
(c) The 50 students represent the population.
(d) The probability any individual in the population will be in the sample is 0.20.

2.3.2 Other Types of Random Sampling

5. A professor is interested in learning more about her students, and she decides that
she is going to draw a sample of 10 students from her population of 50. She knows
she should be drawing some sort of random sample, but doesn’t quite know how to
go about it. She considers the following two sampling designs.

I. Put all the names of students on small pieces of paper, put them into a hat,
and mix them up such that the names are randomly distributed. Then she
Balka ISE 1.09 2.4. EXPERIMENTS AND OBSERVATIONAL STUDIES 5

will pick 10 names without replacement.

II. To ensure equal representation of males and females, she will have separate
hats for males and females. In the frst hat she will mix up the names of the
male students, then draw 5 names without replacement. She will carry out
the same procedure with the females, drawing 5 female names. She will then
pool these names together into a group of 10.

Which of the following statements are true?

(a) Sampling method I results a simple random sample.

(b) Sampling method II results in a stratifed random sample.
(c) Both sampling designs result in very biased samples.

2.4 Experiments and Observational Studies

6. A doctor is investigating the relative e˙ectiveness of two di˙erent procedures that

may prolong life in patients with advanced heart disease. Twenty volunteers are
randomly assigned to the two procedures, 10 to procedure A and 10 to procedure
B. The time until death is recorded.

(a) Is this an experiment or an observational study?

(b) What are the experimental units?
(c) What are the treatments?
(d) What is the response variable?
(e) If it is found that those patients from procedure B tended to live much longer
on average, is it reasonable to conclude that procedure B caused this increase?

7. Researchers are interested in the yield of four types of corn. There are 20 plots of
land available, and the researchers randomly assign the types of corn to the 20 plots
of land (5 plots for each type of corn). After the growing season, the yield (kg) is
measured.

(a) Is this an observational study or an experiment?

(b) What are the treatments?
(c) What is the response variable?
(d) What are the experimental units?

8. Suppose a long-term observational study fnds that children who watch an average
of more than 50 hours of television a week are much more likely to be arrested by
age 18 than those children who watch less television. Based on this study, is it
Balka ISE 1.09 2.5. CHAPTER EXERCISES 6

reasonable to conclude that increased television viewing causes an increase in the

likelihood of arrest?

2.5 Chapter Exercises

9. A study1 investigated fast food consumption among students on a college campus in

the southern United States. The researchers recruited students by posting fyers on
campus, o˙ering to give participants small giveaways and a chance to win a $100 gift
card from the campus bookstore. The 152 student participants completed several
surveys, and one of the surveys involved fast food consumption. In the analysis of
the results of one of the surveys, the researchers found that male students tended
to spend more per month on fast food than female students did.

(a) What is the population of interest?

(b) Is the sample of students a simple random sample?
(c) What types of bias might be present in a study like this?
(d) Would it be possible to conduct a simple random sample of students at this
university to investigate fast food consumption?

10. A study2 investigated physical characteristics of two species of lizard (Phryno-

cephalus frontalis and P. versicolor ) found in a region of Inner Mongolia. The
researchers were interested mainly in the di˙erences between males and females
within each species. Researchers captured these lizards by hand or by noose, and
measured various physical characteristics.

(a) What are the populations of interest?

(b) Did the researchers use a simple random sampling design?
(c) What types of bias might be present in this study?

11. A study3 investigated a possible e˙ect of a vitamin C supplement on the frequency

and duration of colds. One thousand volunteers were randomly assigned to two
groups. One group received a vitamin C supplement of 1g per day, the other group
received a placebo. The participants then kept track of variables related to cold
frequency and severity for the duration of the study (approximately 2 months).

(a) This experiment was conducted as a double-blind experiment (neither the

1
Heidal et al. (2012). Cost and calorie analysis of fast food consumption in college students. Food and
Nutrition Sciences, 3:942–946
2
Qu et al. (2011). Sexual dimorphism and female reproduction in two sympatric toad-headed lizards,
Phrynocephalus frontalis and P. versicolor (Agamidae). Animal Biology, 61:139–151
3
Anderson, T., Reid, D., and Beaton, G. (1972). Vitamin C and the common cold: a double-blind trial.
Canadian Medical Association Journal, 107:503–508
Balka ISE 1.09 2.5. CHAPTER EXERCISES 7

study participant nor the researchers involved knew whether the participant
was receiving the vitamin C or the placebo, until after the study was com-
pleted and the data was summarized). Why would the researchers set up the
experiment in this way?
(b) The study began with 1000 participants, but 182 people dropped out before
the end of the study. The data for the individuals that dropped out was thrown
out and the analysis was based on the remaining 818 participants. How might
this bias the results?

12. Researchers investigated a possible e˙ect of exercise on tumour growth in rats.

Thirty rats were injected with cancerous cells that would develop into a tumour.
These rats were then randomly assigned to one of two groups (15 to each group).
Rats in one of the groups were forced to run on a treadmill for 30 minutes per day.
Rats in the other group had no forced exercise. After one month, the rats were
sacrifced and the weight of the tumour was measured.

(a) What is the response variable?

(b) What is the explanatory variable?
(c) On one of the days, a rat caught his foot on the back of the treadmill and was
badly injured. The rat had to be sacrifced and omitted from the experiment.
How might this bias the results? Does this render the results of the study
invalid?

13. Consider the following list of Central American countries:

• Belize
• Costa Rica
• El Salvador
• Guatemala
• Honduras
• Nicaragua
• Panama

(a) Number the countries from 1–7, then pick a simple random sample of size 3
from this group using the following random numbers:
98797 18334 57628 33215 32849.
What countries are selected? (Hint: We would ignore the frst 2 numbers in
the list (9,8), since no country is labelled with a 9 or an 8, and the frst country
selected would be Panama (7).)
(b) Suppose instead of the above, we randomly pick a letter from A-Z using a ran-
dom number generator, then include the country in our sample if the country’s
name begins with that letter. If we continued in this fashion until 3 countries
were selected, would that sample represent a simple random sample from the
Balka ISE 1.09 2.5. CHAPTER EXERCISES 8

population?

14. None of the sampling methods in this question are perfect, but which one would
yield the most representative sample from its population? (This question requires
a little thought, and the correct answer is subject to debate.)

(a) A student wants to know the class average on a midterm exam she recently
wrote. At the lecture following the midterm, Friday at 8:30, she arrives to
class early and asks ten students in the lecture room what their mark was on
the test. She calculates her estimate of the class average based on what these
10 students say they scored on the test.
(b) A student wants to investigate the average yearly income of the parents of
university students. While in line to obtain his student loan, he asks 10 people
in line with him what their parents’ yearly income is, and takes an average of
the responses.
(c) A teaching assistant wants to know what proportion of students in a class are
male. This particular teaching assistant is a little lazy, and simply looks at
the frst 10 students on a class list that is sorted alphabetically by last name,
and fnds the proportion of these students that are male.

15. Researchers are interested in estimating the proportion of full-time female graduate
students at a university that have a job outside of the university. Surveys are sent
out to all full-time female graduate students, and 67 surveys are completed and
returned. Of these 67 students, 19 have jobs outside of the university.

(a) What is the population of interest?

(b) What is the sample?
(c) What is the parameter of interest?
(d) What is the statistic that estimates this parameter?
(e) Is this an experiment?
(f) Is this a simple random sample?
(g) Will this sample be representative of the population as a whole?

16. Suppose that a person is interested in estimating the average weight of black bear
gall bladders in Ontario. They have a circle of hunting friends, and they fnd
that their next 3 kills, the bears have an average gall bladder weight of 22 grams.
Suppose that, in reality, the average weight of Ontario black bear gall bladders is
24 grams. (Please don’t read into this question as support for bear hunting or the
sale of parts, because it’s not.)

(a) What is the population of interest?

(b) Is 22 grams the value of a parameter or a statistic?
(c) Is 24 grams the value of a parameter or a statistic?
Balka ISE 1.09 2.5. CHAPTER EXERCISES 9

(d) The sampling method could best be described as:

i. Simple random sampling.
ii. Stratifed random sampling.
iii. Voluntary response sampling.
iv. Something else.

17. Suppose a botanist has grown 11 sunfowers and wishes to randomly select 2 of
them for further study. She orders the 11 sunfowers from largest to smallest, and
numbers them from 02 to 12 (in that order). She doesn’t have access to a computer
or a random number table, but she does have a pair of dice available. She rolls the
two dice, gets the sum of the numbers on the two faces, and uses that number to
select the frst sunfower for her sample. She rolls again to randomly select a second
sunfower. Note that the chances of rolling the various possible sums are di˙erent
(the probability of rolling numbers that sum to 2 is 1/36, whereas the probability
of rolling numbers that sum to 7 is 6/36).

Do the two sunfowers that the botanist selected represent a simple random sample
from the population of 11 sunfowers?

18. Suppose a pro-gun organization in the U.S. sends a survey to 1000 of its members,
asking “Do you believe that all illegal immigrants should be deported back to their
country of origin?”. 912 members return the survey, 59% responding “Yes”. This
survey is then used as evidence that 59% of Americans believe that illegal immi-
grants should be deported. (The survey results given in this question are fctitious
and are not meant to be accurate.)

Which of the following statements are true? (check all that are true—there may be
more than one correct statement)

(a) Since only 912 of the 1000 people who were sent the survey returned it, the
survey gives absolutely no evidence about the feelings of members of this or-
ganization toward illegal immigrants.
(b) The biggest problem with this survey is that it is a mail-in survey, which can
cause a delay of several weeks.
(c) The sample size is far too small to be meaningful.
(d) The sample is not representative of the population of interest.

19. A principal wants to estimate the average IQ of third graders in her school. The
frst 10 third graders on the alphabetical list are selected, and their average score
on an IQ test is found to be 97.2.

(a) What is the population of interest?

(b) Is the number 97.2 a parameter, or a statistic?
(c) Is the principal’s sample a simple random sample?
Balka ISE 1.09 2.5. CHAPTER EXERCISES 10

20. Suppose a very conservative media outlet in the U.S. asks a survey question of its
viewership, “Do you approve or disapprove of the way the president is handling his
job as President of the United States?”. This poll has 1500 respondents, 72% of
this sample disapproving of the way the president is handling his job as president.
Suppose this statement is then used to imply that 72% of Americans disapprove of
the way the president is handling his job.

Which of the following statements are true?

(a) The 1500 respondents represent a simple random sample.

(b) The wording of the question is very misleading and would inevitably result in
a biased sample.
(c) The number 72% is a parameter.
(d) Since only 1500 people were sampled from a population of over 300 million,
the sample size is far too small to be meaningful.

21. A study investigated the e˙ect of a regular dose of aspirin on the risk of heart
attack. Physicians randomly assigned approximately 11,000 people to a group that
took an aspirin every second day, and 11,000 to a control group that took a placebo
instead. After several years the subjects in the aspirin group were found to have a
lower risk of heart attack.

(a) Is this an experiment or an observational study?

(b) What is the response variable?
(c) What is the explanatory variable?
(d) Is it necessary to have exactly half the subjects in one group and half in the
other?
(e) True or False: The heart attack variable is confounded with the aspirin vari-
able.

22. Consider again the information in Question 21. Suppose that instead of carrying
out an experiment, the researchers sent out a survey to physicians, asking them if
they had taken aspirin regularly over the past 5 years, and if they had had a heart
attack in that time.

(a) If the data from this survey established a link between aspirin use and a lower
rate of heart attack, would the conclusions be as strong as those found in
Question 21?
(b) What sort of biases are possible in this survey?

23. Researchers wish to investigate the e˙ect of negative reinforcement (a mild electric
shock) on the ability of people to concentrate. Twenty people volunteer for this
investigation. Ten are randomly assigned to a shock group, and ten are assigned to
a control group. Each individual plays a memory game, involving 100 tiles laid face
Balka ISE 1.09 2.5. CHAPTER EXERCISES 11

down on a table. The 100 tiles each have a picture on the face-down side. There
are 50 pairs of identical pictures, and the individual must attempt to fnd matching
pairs. Individuals assigned to the shock group get an electric shock whenever their
attempted match is wrong. Individuals in the control group do not get a shock.
Each individual is allowed 100 attempts at fnding matching pairs of tiles, and the
number of matching pairs they fnd is recorded.

Which of the following statements are true?

(a) There are 20 treatment groups.

(b) This is an observational study.
(c) Even if there is strong evidence of a di˙erence between the two groups, there
will not be evidence of a causal e˙ect.
(d) The response variable is the voltage of the electric shock.

24. A company is concerned that there are too many injuries on one of their assembly
lines. They decide that they may need to make certain workers wear a type of
harness to prevent falls and injuries. They are concerned that the harness may
cause the workers to work more slowly and have longer task completion times.
Fifteen workers volunteer for a study to investigate this. Eight of these workers are
randomly assigned to the harness-wearing group, and the remaining seven workers
will not wear a harness. The workers are observed for two weeks and their task
completion times are recorded.

(a) What is the response variable?

(b) What is the explanatory variable?
(c) Is this an experiment or an observational study?
(d) Can this type of study help determine if there is a causal link between wearing
the harness and slower times?
(e) Why did the company randomly assign the workers to the groups, and did not
allow the workers to pick which group they wanted to be in?

25. You want to perform an experiment to investigate the e˙ect of di˙erent stimuli on
the time it takes for rats to complete a maze. You decide to have 3 groups of rats.
Group 1 is a control group, rats in Group 2 will receive a piece of cheese if they
complete the maze within a certain time, and rats in Group 3 will get a mild (but
painful) electric shock every 3 seconds until they complete the maze. You have 9
rats available for the experiment, and you (wisely) decide to randomly assign 3 rats
to each group. You have named the rats:
Pete, Tom, Jerry, Alphonse, Renaldo, Bo, Little Pete, Huey, Eddie.

Number the rats from 1–9 using a single digit. Then randomly assign the rats to
the groups by using the following random numbers.
Balka ISE 1.09 2.5. CHAPTER EXERCISES 12

Random Number Line: 33696 91544 76248

(The frst 3 chosen go to the control group, the next 3 to the cheese group, and the
remainder are the unlucky ones in the electric shock group.)

(a) Which rats go in which group?

(b) Little Pete is your favourite rat, and you do not want to see him shocked. You
decided to put him in the cheese group, and randomize the others. Why is
this not acceptable from a statistical point of view?

26. Suppose a new yoga centre conducts a survey of 1000 women. They fnd that those
who practice yoga regularly are much healthier than those that do not. They use
this fnding in their advertising, stating that their survey shows that “regularly
practicing yoga can improve women’s health”. What, if anything, is wrong with
that statement?

27. Which one of the following is the most reasonable defnition of a lurking variable?

(a) An unmeasured variable that is related to both the response and explanatory
variables.
(b) An unmeasured variable that is related to the response variable, but has no
relationship with the explanatory variable.
(c) A measured variable that a˙ects the response, but is not related to the ex-
planatory variable.
(d) A variable that has a strong e˙ect on the mean of a data set, but does not
a˙ect the median.
(e) An explanatory variable that causes changes in the response variable. Exper-
iments can give strong evidence of a causal relationship between variables.

28. Researchers are interested in assessing the e˙ect of four di˙erent drugs on patients
with high blood pressure. Two hundred people with high blood pressure volunteer
for an experiment. The 200 people are randomly assigned to the di˙erent drugs,
50 to each drug. After 12 weeks on the drug, the reduction in blood pressure is
recorded.

(a) What is the response variable?

(b) What is the explanatory variable?
(c) Is this an experiment, or an observational study?
(d) Can this type of study give strong evidence of a causal link between the drug
that is taken and reduced blood pressure?

29. An insurance company lets new policy purchasers choose whether they wish to
purchase “accident forgiveness”, which allows the driver to avoid a premium increase
in the event of an at-fault accident. Out of 200 new purchasers, 40 choose to
Balka ISE 1.09 2.5. CHAPTER EXERCISES 13

purchase accident forgiveness. It is found that after one year, 8 of the 40 who
purchased accident forgiveness got into an at-fault accident, whereas only 4 of the
160 that did not purchase it got into an at-fault accident. Which of the following
statements are true? There may be more than one true statement; check all that
are true.

(a) This is an observational study, not an experiment.

(b) There is very strong evidence that purchasing accident forgiveness causes more
drivers to get into an at-fault accident.
(c) This study shows that there is defnitely no relationship between purchasing
accident forgiveness and getting into an accident.

30. Dental researchers are investigating the strength of two new bonding agents, and
comparing their bond strength to that of the commonly used (standard) bonding
agent. They run an experiment in which the 3 compounds are tested on extracted
teeth. There are 60 extracted teeth available for the experiment. The 60 teeth are
randomly assigned to the 3 bonding compounds, 20 to each compound. The re-
searchers attach a dental appliance to each tooth, and the bond strength is recorded.

(a) What are the treatments?

(b) What is the explanatory variable?
(c) Is this an observational study, or an experiment?
(d) Can this type of study help establish if a there is a causal link between the
bonding agent and the strength of the bond?

31. Consider the following scenarios. In each scenario determine whether it would be
best investigated with an experiment, a survey, or an observational study that is
not a survey.

(a) You are conducting a study investigating drivers in fatal single-vehicle acci-
dents. You want to determine whether male drivers are more likely than female
drivers to have alcohol in their system.
(b) You want to determine whether new male MBA graduates have a higher salary
on average than new female MBA graduates.
(c) You want to determine which of two commonly used headache medications
results in the fastest pain relief.
(d) You want to determine if there is a relationship between cocaine use and
marital infdelity.
(e) You want to investigate whether men who commit suicide tend to use a frearm
more frequently than women who commit suicide.

32. A veterinarian is developing a new experimental surgical technique for dogs that
have a certain type of cancer. The surgery is very risky, and many of the dogs die
soon after surgery. But the veterinarian believes that most dogs would live longer
Balka ISE 1.09 2.5. CHAPTER EXERCISES 14

after the experimental surgery than if they had the standard surgery. She o˙ers
dog owners a choice of the two methods. She fnds that those dogs that undergo the
new experimental surgery live a shorter time on average than those that undergo
the standard surgery.

Disheartened, she decides to carry out an experiment. Dogs in the experiment are
randomly assigned to one of the two surgical methods. In the experiment, it is
found that dogs that undergo the experimental surgery live much longer on average
than dogs that undergo the standard surgery.

(a) If you have a dog with bone cancer, and your sole objective is to maximize
their length of life after surgery, which surgery method should you choose?
i. The new experimental surgery.
ii. The standard surgery.
(b) Which one of the following is the most important lurking variable that would
a˙ect the interpretation of the results of the observational study?
i. The medical condition of the dog.
ii. The size of the dog.
iii. The income of the dog owners.
iv. The IQ of the veterinarian.
v. The breed of the dog.
Chapter 3

Descriptive Statistics

J.B.’s strongly suggested exercises: 1, 4, 7, 8, 9, 10, 11, 16, 17, 20, 21, 22, 24, 25, 36, 46,
49

3.1 Introduction

3.2 Plots for Qualitative and Quantitative Variables

3.2.1 Plots for Qualitative Variables

1. A Finnish study1 investigated a possible association between the gender of convicted

murderers and their relationship to the victim. A random selection of 91 female
murderers and a random selection of 91 male murderers were obtained, and the
results are illustrated in Figure 3.1.

(a) For male murderers, summarize the distribution of the relationship between
the murderer and the victim.
(b) Describe the observed di˙erences in the distributions of the relationship to the
victim between male and female murderers.
(c) Can we be certain that the observed di˙erences in the samples of male and
female murderers refect the true di˙erences in the population distributions?
(d) Sketch a pie chart to illustrate the distribution of the relationship to the vic-
tim for male murderers. (Sketch a rough plot – it does not need to be very
1
Häkkänen et al. (2009). Gender di˙erences in Finnish homicide o˙ence characteristics. Forensic Science
International, 186:75–80.

15
Balka ISE 1.09 3.3. NUMERICAL MEASURES 16

0.6
Acquaintance

Relative Frequency
0.5 Partner
Family
0.4 Stranger
Gender Male Female Total 0.3

Acquaintance 61 37 98 0.2
Partner 22 32 54
0.1
Family Member 4 18 22
0.0
Stranger 4 4 8 Male Offenders Female Offenders

Total 91 91 182

Figure 3.1: Gender of murderer and their relationship to the victim in Finnish murders.

accurate.)

3.2.2 Plots for Quantitative Variables

3.3 Numerical Measures

3.3.1 Summation Notation

2. Suppose we have the following sample data set: 8, 14, 22, −5

(a) What is the value of x3 ?

X3
(b) What is xi ?
i=1
(c) What is x?
P
P i2
(d) What is xi ?

3.3.2 Measures of Central Tendency

3.3.2.1 Mean, Median, and Mode

3. Suppose we have a sample of 5 observations: 1, 5, 2, −3, 987.

(a) What is the mean?

(b) What is the median?
(c) What is the mode?
Balka ISE 1.09 3.3. NUMERICAL MEASURES 17

(d) If the extreme value is removed, what is the mean?

(e) If the extreme value is removed, what is the median?

4. A random sample of 4 Canadian male newborns revealed the following birth weights,
in grams:
2870, 2620, 3120, 3620

(a) What is the mean birth weight?

(b) What is the median birth weight?
(c) What are the units of the mean?
(d) What are the units of the median?
(e) In this situation, which is the more appropriate measure of central tendency,
the mean or the median?

3.3.2.2 Other Measures of Central Tendency

5. What would be an advantage of using the trimmed mean instead of the untrimmed
mean? What would be a disadvantage?

3.3.3 Measures of Variability

6. Consider the following sample of 4 observations: 18, 8, 3, 17.

(a) What are the deviations?

(b) What is the sum of the deviations?

7. A random sample of 4 Canadian male newborns revealed the following birth weights,
in grams:
2870, 2620, 3120, 3620

(a) What is the range?

(b) What is the mean absolute deviation?
(c) What is the variance?
(d) What is the standard deviation?
(e) What are the units of the variance?
(f) What are the units of the standard deviation?

8. Create a 4 number sample data set, where all numbers lie between 0 and 500
(inclusive, and repeats are allowed), such that the standard deviation is as large as
possible.
Balka ISE 1.09 3.3. NUMERICAL MEASURES 18

(a) What is your data set?

(b) What is the standard deviation of your data set?

9. Create a 4 number sample data set, where all numbers lie between 0 and 500
(inclusive, and repeats are allowed), such that the standard deviation is as small as
possible.

(a) What is your data set?

(b) What is the standard deviation of your data set?

10. Which of the following statements are true? There may be more than one correct
statement; check all that are true.

(a) The standard deviation can be greater than the variance.

(b) The standard deviation can be negative.
(c) The standard deviation can be less than the mean.
(d) The standard deviation can be less than the third quartile.
(e) The standard deviation is always less than the average distance from the mean.

3.3.3.1 Interpreting the standard deviation

11. Figure 3.2 illustrates scores on a diÿcult statistics test. The scores have a mean of
22.4 and a standard deviation of 7.2. (The maximum possible score on the test was
40.)

(a) Would the empirical rule apply to this data? Why or why not?
(b) What would the empirical rule tell us about the proportion of observations
that are within 7.2 of 22.4?
(c) What would the empirical rule tell us about the proportion of observations
that are within 14.4 of 22.4?
(d) What would the empirical rule tell us about the proportion of observations
that are within 21.6 of 22.4?

12. Figure 3.2 illustrates scores on a diÿcult statistics test. The scores have a mean of
22.4 and a standard deviation of 7.2.

(a) Would Chebyshev’s theorem apply to this data? Why or why not?
(b) What would Chebyshev’s theorem tell us about the proportion of observations
that are within 7.2 of 22.4?
(c) What would Chebyshev’s theorem tell us about the proportion of observations
that are within 14.4 of 22.4?
(d) What would Chebyshev’s theorem tell us about the proportion of observations
that are within 21.6 of 22.4?
Balka ISE 1.09 3.3. NUMERICAL MEASURES 19

60

Frequency
40

Frequency

20

0
0 10 20 30 40
Scoretemp2
on Test
.
Figure 3.2: Scores on a hard test (x̄ = 22.4, s = 7.2).

13. Consider the histogram given in Figure 3.3.

(a) Would the empirical rule apply to this data?

(b) Would Chebyshev’s theorem apply to this data?

Figure 3.3: A distribution that is skewed to the right.

3.3.3.2 Why divide by n − 1 in the sample variance formula?

14. Why do we divide by n − 1 in the sample variance formula?

15. Suppose we have a sample of size n = 87, and the population mean and variance
are unknown. How many degrees of freedom are there for estimating the variance?
Balka ISE 1.09 3.3. NUMERICAL MEASURES 20

3.3.4 Measures of Relative Standing

3.3.4.1 Z-scores

16. A random sample of 4 Canadian male newborns revealed the following birth weights,
in grams:
2870, 2620, 3120, 3620
In Exercise 7, we found that for these four observations: x̄ = 3057.5 and s =
426.9563.

(a) What are the 4 z-scores?

(b) What is the mean of the 4 z-scores?
(c) What is the standard deviation of the 4 z-scores?
(d) If a newborn male baby had a z-score of 4.6, what does that tell us about the
baby’s weight?
(e) If a newborn male baby had a z-score of −0.4, what does that tell us about
the baby’s weight?

17. Todd has always had a dream of becoming a medical doctor. After doing well in
an introductory statistics course, Todd decides to write the MCAT. His score on
the test corresponded to a z-score of 3.0. Suppose that scores on the test have a
distribution that is mound-shaped (approximately normal). Which of the following
statements are true?

(a) The z-score is a unitless quantity.

(b) Todd’s score was 3 standard deviations greater than the mean score.
(c) Todd scored worse than approximately 1/3 of the test writers.
(d) Todd’s score was better than average.

3.3.4.2 Percentiles

18. A sample of 8 boxes of Kellogg’s All Bran was collected at a grocery store. The
boxes had a nominal weight of 675 grams. The weight (in grams) of the cereal in
each box was recorded, with the following results:

684, 684, 686, 691, 691, 686, 691, 684

(The weights were recorded after discarding the box and the bag, so they represent
the weight of only the cereal.)
Balka ISE 1.09 3.4. BOXPLOTS 21

(a) Use the method outlined in the text to calculate the 80th percentile of the
weights.
(b) Use the method outlined in the text to calculate the 25th percentile of the
weights.

19. The 90th percentile of heights of adult females in the United States is closest to
which one of the following?

(a) 90 cm.
(b) 122 cm.
(c) 143 cm.
(d) 171 cm.
(e) 200 cm.

3.4 Boxplots

20. Consider the boxplots in Figure 3.4, representing 3 di˙erent samples.

80
60
40
20
0

A B C

Figure 3.4: 3 boxplots.

(a) What is the median of sample C?

(b) What is the range (Maximum − Minimum) of sample C?
(c) How many outliers are there in the entire plot (all samples combined).
(d) What is the 25th percentile of sample C?

21. Qu et al. (2011) investigated physical characteristics of the lizard Phrynocephalus

frontalis. In one part of the study, the researchers compared the tail lengths of
males and females of this species. Figure 3.5 illustrates the distributions of tail
length for 44 female and 22 male lizards that were captured in the wild.
Balka ISE 1.09 3.5. LINEAR TRANSFORMATIONS 22

75

Tail Length (mm)

70

65

60

Females Males

Figure 3.5: Boxplots of tail lengths of 44 female and 22 male lizards of the species P.
frontalis.

(a) What is the 75th percentile of tail lengths for the sample of male lizards?
(b) What is the 75th percentile of tail lengths for the sample of female lizards?
(c) Summarize the main di˙erences and similarities between males and females
for these samples of tail lengths.

3.5 Linear Transformations

22. A certain sample has a mean of 100, a median of 90, and a standard deviation of
20. Suppose that each observation has 5 added to it, and the result is multiplied
by 20.

(a) What is the mean of the new values?

(b) What is the variance?
(c) What is the standard deviation?
(d) What is the median?

23. Suppose we have a sample of 7 values, and we calculate the usual summary statistics.
If we add 5,000 to each of these 7 sample values, then recalculate the summary
statistics, which of the following quantities would change?

mean, median, Q3 , IQR, variance, standard deviation

Balka ISE 1.09 3.6. CHAPTER EXERCISES 23

3.6 Chapter Exercises

3.6.1 Basic Calculations

24. Consider the following sample of 4 numbers: 52, −2, 8, 107.

(a) What is the mean?

(b) What is the median?
(c) What is the variance?
(d) What is the standard deviation?
(e) What is the range?

25. Consider the following sample of 7 observations:

3.1, 8.2, 9.6, −1.7, 8.4, 8.8, 21.1

(a) What is the mean?

(b) What is the median?
(c) What is the third quartile (Q3 )?
(d) What is the interquartile range?
(e) What is the variance?
(f) What is the standard deviation?
(g) Are any of these observations outliers according to the 1.5*IQR rule?

26. Consider the following sample data set: 32, 36, 1, 4, 89, 18.

(a) What is the mean?

(b) What is the median?
(c) What is the value of Q1 ?
(d) What is the value of the interquartile range?
(e) What is the variance?
(f) What is the standard deviation?

27. Consider the following sample data set: −4, 6, −12, 14, 742

(a) What is the mean of the full data set?

(b) What is the median of the full data set?
(c) What is the standard deviation of the full data set?
(d) What is the variance of the full data set?
(e) If the outlier is discarded, what is the mean?
(f) If the outlier is discarded, what is the median?
(g) If the outlier is discarded, what is the standard deviation?
Balka ISE 1.09 3.6. CHAPTER EXERCISES 24

28. Consider the following computer output, illustrating a stemplot.

Decimal point is at the colon

-1 : 7110
-0 : 76632
0 : 03335
1 : 01389
2 : 4

(a) What is the mean?

(b) What is the median?
(c) What is the variance?
(d) What is the standard deviation?
(e) What is the range?

29. Consider the following split-stem stemplot: (The decimal is at the colon. You
should be able to recognize that the largest observation is 5.8.)

0 : 0000000000000000000111111222222233334444444
0 : 557777888
1 : 001111224444
1 : 5555778999
2 : 011223
2 : 55566689
3 : 0
3 : 5678
4 :
4 : 7789
5 :
5 : 668

(a) Is the mean greater than the median?

(b) Does the distribution show any skewness? If so, in what direction is it skewed?
(c) What is the value of the di˙erence between the ffth and frst values in the
fve-number summary?

30. Consider the following stemplot:

Balka ISE 1.09 3.6. CHAPTER EXERCISES 25

-4 : 975
-3 : 99410
-2 :
-1 : 6
-0 :
0 :
1 : 1
2 : 13
3 : 249

(The decimal place is at the colon.)

(a) What is the mean?

(b) What is the median?
(c) If the decimal place was shifted one place to the left for all the observations
(e.g. the smallest observation was −0.49), which of the following statistics
would change?

Mean, median, Q1 , Q3 , IQR, standard deviation, variance

3.6.2 Concepts

31. Consider the bimodal frequency histogram for a variable x, illustrated in Figure 3.6.

50

40
Frequency

30

20

10

0
-20 -10 0 10 20 30 40
x
Figure 3.6: A bimodal frequency histogram.

(a) What is the approximate value of the median?

(b) What is the approximate value of Q1 ?
(c) What is the approximate value of Q3 ?
(d) Are there any outliers?
Balka ISE 1.09 3.6. CHAPTER EXERCISES 26

(e) True or false: The mean and median would be close in value.
(f) One of the boxplots in Figure 3.7 represents this data. Which one?

40 40
20

30 30

15
20 20

10 10 10

0 0
5

-10 -10

A B C

Figure 3.7: Which boxplot corresponds to the histogram?

32. Suppose we have a sample of 5 observations. We forget the data values, but we
know the frst 4 deviations from the mean (xi − x̄) are:

−2.3, 3.7, 1.2, −0.7

(a) What must the value of the ffth deviation be?

(b) What was the sample mean?
(c) What is the standard deviation?

33. If the smallest 4 values in a sample with n = 5 are equal, and the range is equal to
10, what is the value of the variance?

34. Consider the frequency histogram shown in Figure 3.8. Which of the following
statements are true?

(a) The median is less than 4.

(b) The mean is less than the median.
(c) The IQR is greater than 5.
(d) All the numbers in the fve number summary are positive.
(e) This distribution is approximately symmetric.
(f) The value of Q1 −Median would have the same absolute value as Q3 −Median.
(g) There are no obvious outliers.

35. Consider the boxplots in Figure 3.9.

(a) Which one of the boxplots corresponds to the histogram in Figure 3.8?
Balka ISE 1.09 3.6. CHAPTER EXERCISES 27

40

30

Frequency
20

10

0
0 1 2 3 4 5
x
Figure 3.8: A frequency histogram.
7

5 5
6

4 4
5

3 3
4

2 2 3

1
1 2

0
A B C

Figure 3.9: Which boxplot represents the histogram data?

(b) Which boxplot has the largest median?

(c) Which boxplot has the largest interquartile range?
(d) Which boxplot contains the largest number of outliers?

36. Suppose we have a sample of 12 observations, and we calculate the summary statis-
tics for these values.

(a) If we add 100,000 to the largest observation, which of the following statistics
would change?
Mean, median, variance, standard deviation, interquartile range
(b) If we add 100,000 to the largest observation, and subtract 100,000 from the
smallest observation, which of the following statistics would change?
Mean, median, variance, standard deviation, interquartile range

37. Suppose there is a data set of 100 observations that has a standard deviation equal
to 0. Which of the following statements are true?

(a) All the numbers in the 5 number summary must be equal.

Balka ISE 1.09 3.6. CHAPTER EXERCISES 28

(b) There are no negative deviations.

(c) The variance and standard deviation are equal.
(d) The mean and median are equal.
(e) If a single observation from this data set had a positive constant added to it,
then the IQR, minimum, and median would not change.

38. Suppose we have a sample data set with 4 observations, where x̄ = 50 and s = 0.
Suppose we add one more observation to this data set, and we recalculate the
mean and standard deviation. If the added observation equals 50.0, what will the
standard deviation of the 5 number data set be?

39. Which of the following statements are true?

(a) The standard deviation is always greater than the median.

(b) The variance is always greater than the standard deviation.
(c) If a distribution is skewed to the left, then the mean will always be less than
the standard deviation.
(d) If all values in a sample are within 2.0 of each other, then the standard devi-
ation will be less than 2.0.
(e) The median can equal the third quartile (Q3 ).

40. Which of the following statements are true?

(a) If the median of a distribution equals the minimum, then the distribution can’t
be skewed to the left.
(b) If a variable x is measured in seconds, then the units of the variance are
seconds2 .
(c) The 12.5th percentile is exactly half the value of Q1 .
(d) All of the values in the fve-number summary are always positive.
(e) Large outliers have little e˙ect on the standard deviation, but can have a
strong e˙ect on the variance.

41. Which of the following statements are true?

(a) For right-skewed distributions, the median is greater than the mean.
(b) For symmetric distributions, the standard deviation and mean are equal.
(c) For mound-shaped distributions, approximately 50% of observations lie within
5 standard deviations of the mean.
(d) For any distribution, the interquartile range is equal to 50.
(e) All observations that are less than Q1 would have negative z-scores. (Careful!)
(f) If the range of a data set is equal to 0, then the variance and IQR will also
equal 0.

42. Consider the following sample data:

Balka ISE 1.09 3.6. CHAPTER EXERCISES 29

−2, −5, 12, 17, 15, 15, 15, 15, 15, 15, ?, ?

(The question marks represent two missing values.)

(a) What is the value of the median for the entire sample (including the missing
values)? If it is impossible to determine, say so.

(b) What is the value of the mean for the entire sample (including the missing
values)? If it is impossible to determine, say so.

(c) What is the value of the variance for the entire sample (including the missing
values)? If it is impossible to determine, say so.

43. A certain data set has a mean of 15 and a standard deviation of 5. After a linear
transformation, the new standard deviation was found to equal 1, but the mean
was unchanged.

(a) What possible linear transformations result in these new values (there are two
possibilities).
(b) Suppose the appropriate transformation is the one where b is positive. If a
value in the old data set is 10, what is its new value?

44. In a sample of size n = 3, the standard deviation is equal to 100. Two of the
observations are 152 and 252. The third observation is one of two possible values.
What are these two possible values?

3.6.3 Applications

45. The histogram in Figure 3.10 represents fnal grades for students in an introductory
statistics course. (This is real data from one of your author’s courses.)

120
Frequency

80

40

0
20 40 60 80 100
Final Grade

Figure 3.10: Final grades in an introductory statistics course.

Balka ISE 1.09 3.6. CHAPTER EXERCISES 30

(a) What is the approximate value of the median?

(b) What is the approximate value of Q1 ?
(c) What is the approximate value of the interquartile range?
(d) Would the mean be greater or less than the median?
(e) What is the approximate value of the standard deviation?

46. Figure 3.11 is a histogram of grades on a major assignment in a large introductory

statistics course. (This is real data from one of your author’s courses.) Of the 206
students in the course, twelve students did not hand in the assignment and received
a grade of 0. The average grade for all students was 18.8 out of 25.
25
20
Frequency
15
10
5
0

0 5 10 15 20 25
Grade on Assignment

Figure 3.11: Grades on a major assignment.

(a) Describe the distribution.

(b) Reporting the mean grade for all students might be slightly misleading in this
setting. Suggest a better numerical summary of this data.

47. The insect Megamelus scutellaris can feed on the sap of the water hyacinth, and
it has been used as a method of biocontrol for this invasive plant species. M.
scutellaris mates on the water hyacinth, and females create oviposition scars on the
plant, laying one or more eggs in each scar. Sosa et al. (2005) investigated several
aspects of the reproductive cycle of this insect. In one aspect of this study, the
number of eggs laid per scar was measured. The observed number of eggs per scar
for 359 scars is given in Table 3.1.

1 egg/scar 2 eggs/scar 3 eggs/scar 4 eggs/scar

142 194 9 14

Table 3.1: Observed number of M. scutellaris eggs per scar.

(a) What is the mean number of eggs per scar?

(b) What is the median number of eggs per scar?
(c) What is the standard deviation of the number of eggs per scar?
Balka ISE 1.09 3.6. CHAPTER EXERCISES 31

48. As part of an investigation by Fink et al. (2007), 32 male student volunteers at a

university in Germany had their hand grip strength measured (in kilograms force
(kgf)). The results are illustrated in Figure 8.1.

Hand Grip Strength (kgf)

60

6 55
Frequency

50
4
45

2 40

35
0
30 35 40 45 50 55 60 65
Hand Grip Strength (kgf)

Figure 3.12: Plots of the grip strengths of 32 male student volunteers.

(a) Describe the distribution of the grip strengths.

(b) Estimate the 90th percentile of the grip strengths in this data set.
(c) Estimate the 90th percentile of the grip strengths of adult Germans.
(d) For this data, is the mean or the median the more appropriate measure of
central tendency?
(e) Give rough estimates of the mean and median.
(f) Give a rough estimate of the standard deviation.

49. Tantius et al. (2014) investigated tensile properties of human umbilical cords. In
one part of the study, 23 human umbilical cords were stretched until they broke, and
their elongation (% increase in length) at the breaking point was recorded. (The
breaking point of umbilical cords is of interest in forensic science, as mothers accused
of killing a newborn might claim that an umbilical cord broke during childbirth and
the newborn baby bled to death.) The elongation percentages are illustrated in
Figure 3.13.
120
100
Elongation (%)
80
60
40
20

Figure 3.13: Boxplot of elongation percentage at breaking point.

(a) Describe the distribution of the elongation percentages.

Balka ISE 1.09 3.6. CHAPTER EXERCISES 32

(b) The mean elongation percentage of these 23 umbilical cords is 36.6. Estimate
the mean if the outlier were to be removed from the calculations.
(c) Should the outlier be omitted from the analysis?
(d) The study actually involved 25 umbilical cords, but in two cases the machine
did not recognize the break and thus did not record the elongation at the break-
ing point. These data points were omitted in this plot and in the calculation
of the mean. How might this distort the results?

50. Harlioglu et al. (2012) investigated several characteristics of crayfsh in a freshwater

lake in Turkey. In this study, twenty-fve adult male Astacus leptodactylus crayfsh
were sampled, and several variables were measured. One of the variables was body
weight (in grams). The weights are illustrated in Figure 3.14.
7

40
6

35
5

Weight (grams)
Frequency
4

30
3

25
2
1

20
0

15 20 25 30 35 40 45
15

Weight (grams)

Figure 3.14: Weights of 25 adult male crayfsh.

(a) Give rough estimates of the mean and median of the 25 weights.
(b) Suppose the researchers wished to estimate the average weight of all adult
male crayfsh in this lake. Would the mean of this sample be a reasonable
estimate of this quantity?

51. The following observations represent lengths, in millimetres, of a sample of 16 He-

liconia bihai fowers.
47.12 46.75 46.81 47.12 46.67 47.43 46.44 46.64
48.07 48.34 48.15 50.26 50.12 46.34 46.94 48.36

Summary statistics: x̄ = 47.5975, median = 47.12, and s = 1.212878.

Suppose we want to change the measurement scale, and have the measurements
represent the number of inches the length is in excess of 40 mm. To carry this out,
we (correctly) subtract 40, then divide the result by 2.54.

(a) What is the mean of the new values?

(b) What is the median of the new values?
(c) What is the standard deviation of the new values?
(d) What is the variance of the new values?
Balka ISE 1.09 3.6. CHAPTER EXERCISES 33

52. A person is weighing rocks in a lab in the United States. They are new to the
United States from Canada, and they have been taking all of the measurements in
kilograms. The sample mean and standard deviation of the weight of the rocks (in
kilograms) were found to be 15.10 and 1.70, respectively.

At this point the person realizes two important points: 1) The scale wasn’t cali-
brated properly, and every measurement in the data set should have 2.00 kilograms
added to it. 2) Their boss wants the mean and the standard deviation in pounds,
not kilograms (1 kg = 2.20 pounds). The person correctly realizes that they must
add 2.00 to all of their measurements, then multiply the result by 2.20. What are
the mean and standard deviation of the correct weights in pounds?
Balka ISE 1.09 3.6. CHAPTER EXERCISES 34
Chapter 4

Probability

J.B.’s strongly suggested exercises: 2, 4, 5, 6, 7, 8, 14, 16, 17, 21, 26, 30, 35, 39, 43, 49,
57, 61

4.1 Introduction

4.2 Basics of Probability

4.2.1 Sample Spaces and Sample Points

1. Two cards are drawn without replacement from a well-shu˜ed deck.

(a) What is the sample space of this experiment?

(b) What are the the sample points? How many sample points are there?
(c) Are the sample points equally likely?
(d) What is the probability of getting a pair of twos?

2. Two balanced six-sided dice are rolled, and the numbers that come up are recorded.

(a) What is the sample space of this experiment?

(b) What are the the sample points? How many sample points are there?
(c) Are the sample points equally likely?
(d) What is the probability of rolling a sum of 2? (Rolling two dice such that the
sum of the values on the up face is 2.)
(e) What is the probability of rolling a sum of 7? (Rolling two dice such that the
sum of the values on the up face is 7.)

35
Balka ISE 1.09 4.3. RULES OF PROBABILITY 36

4.3 Rules of Probability

4.3.1 The Intersection of Events

4.3.2 Mutually Exclusive Events

4.3.3 The Union of Events

4.3.4 Complementary Events

Questions for the intersection of events, mutually exclusive events, the union of
events, and complementary events:

3. For each of the following events, draw a Venn diagram and shade the region that
represents it.

(a) A ∪ Bc
(b) A ∩ Bc
(c) Ac ∪ B c
(d) Ac ∩ B c
(e) (A ∪ B)c

4. Suppose P (A) = 0.6, P (B) = 0.3, and P (A ∩ B) = 0.2.

(a) What is P (Ac )?

(b) What is P (A ∪ B)?
(c) What is P ((A ∪ B)c )?
(d) Are A and B mutually exclusive?
(e) Are Ac and B c mutually exclusive?

5. The following table gives the distribution of blood types in Canada.1

A B AB O
Rh positive 0.36 0.076 0.025 0.39
Rh negative 0.06 0.014 0.005 0.07

Table 4.1: Proportions of Canadians with di˙erent blood types. Every person falls into
one and only one of these 8 categories.

Suppose we randomly select a Canadian. Consider the following events:

1
As given on the Canadian Blood Services website. Accessed January 9, 2014.
Balka ISE 1.09 4.3. RULES OF PROBABILITY 37

A: The person has blood type A.

B: The person has blood type B.
C: The person has blood type AB.
D: The person has blood type O.
E: The person is Rh positive.

(a) Symbolically, what is the event that the person has type O negative blood?
(If a person has type O blood and a negative Rh factor, it is said that they
have type O negative blood.) What is the probability of this event?
(b) What is the probability the person has blood type O or a positive Rh factor?
(c) Which pairs of events are mutually exclusive?
(d) A person with A negative blood can receive a blood transfusion from someone
who also has type A negative blood, or from someone who has type O negative
blood. In symbols, what is the event that a randomly selected Canadian can
donate to someone with type A negative blood? What is the probability of
this event?

Conditional Probability

Independent Events

Questions for conditional probability and independent events:

6. Suppose P (A) = 0.60, P (B) = 0.20 and P (A ∪ B) = 0.65.

(a) What is P (A ∩ B)?

(b) What is P (A|B)?
(c) What is P (B|A)?
(d) Are A and B independent?

7. A balanced six-sided die is about to be rolled once.

• Let A be the event that the number on the top face is greater than or equal
to 4.
• Let B be the event that the number on the top face is even.
• Let C be the event that the number on the top face is a 6.

(a) What is P (B|C)

(b) What is P (C|B)
(c) What is P (C|B c )
(d) What is P ((B ∩ C)|A)?
(e) Which, if any, of the 3 pairs of events (A, B), (A, C), (B, C) are independent?
Balka ISE 1.09 4.4. EXAMPLES 38

The Multiplication Rule

8. Cards are drawn without replacement from a well shu˜ed deck.

(a) If two cards are drawn, what is the probability both cards are hearts?
(b) If two cards are drawn, what is the probability both cards are fves?
(c) If four cards are drawn, what is the probability all four cards are hearts?

9. Two balanced six-sided dice are rolled. (The rolling is done in such a fashion that
di˙erent rolls can be considered independent.)

(a) What is the probability both dice have a four on the up face?
(b) What is the probability that an even number is rolled on the frst die, and a
number greater than four is rolled on the second?

4.4 Examples

10. If P (A) = 0.80, P (B) = 0.40, and P (B|A) = 0.25, what is P (A|B)?

11. Suppose we know the following about events A, B, and C:

P (A) = 0.42, P (B) = 0.44, P (C) = 0.40

P (A ∩ B) = 0.16, P (A ∩ C) = 0.20, P (B ∩ C) = 0.12

P (A ∩ B ∩ C) = 0.04

(a) Draw a Venn diagram and fll in the appropriate probabilities.

(b) What is P (A ∪ B)?
(c) What is P (A ∪ B ∪ C)?
(d) What is P (A|B ∩ C)?

12. The following table is based on a 2009 poll in the U.S., in which 1000 randomly
selected adults were asked if they approved or disapproved of the way Barack Obama
is handling his job as president.

Republicans Democrats Independents

Approve 60 361 139
Disapprove 270 49 121

(a) If a person is randomly selected from this group of 1000, what is the probability
they are Republican?
Balka ISE 1.09 4.5. BAYES’ THEOREM 39

(b) What is the probability they said they approve of the way Obama is handling
his job?
(c) Given a randomly selected person is Republican, what is the probability they
approve?

13. Consider the Venn diagrams shown in Figure 4.1. For each plot, assume that the
area of the square is 1, and that the area of each region is equal to its probability
of occurring. The smaller circle represents event A, and the larger circle represents
event B.

(a) (b)

(c) (d)

Figure 4.1: Venn diagrams for Question 13.

(a) One of the plots was created for the situation where A and B are independent.
Which one is it?
(b) For each plot, state whether P (A|B) = P (A), P (A|B) > P (A), or P (A|B) <
P (A).

4.5 Bayes’ Theorem

14. An airport security oÿcial tests an experimental electronic sniÿng device that can
detect explosive materials. All passengers are scanned by this device, which sounds
an alarm if it detects explosive material. The device is not perfectly reliable. It will
falsely sound an alarm 2% of the time if the person being scanned is not carrying
explosive material It will also fail to sound an alarm 8% of the time if the person
being scanned is carrying explosive material.
Balka ISE 1.09 4.6. COUNTING RULES: PERMUTATIONS AND COMBINATIONS 40

Assume that 1 in 10,000 passengers carries explosive material. If the device scans
a randomly selected passenger:

(a) What is the probability the alarm sounds?

(b) If the alarm sounds, what is the probability the person is actually carrying
explosive material?

15. A small regional airport is serviced by 2 commercial airlines. Caribou Air accounts
for 65% of commercial fight arrivals. Arrivals from this airline are late 22% of the
time. Polar Express Air accounts for 35% of commercial fight arrivals. Arrivals
from Polar Express Air are late 42% of the time.

(a) If a commercial fight arrival is randomly selected, what is the probability it

is late?
(b) Given a randomly selected commercial fight arrival is late, what is the prob-
ability the fight is a Caribou Air fight?

4.6 Counting rules: Permutations and Combinations

16. (a) At the start of a Texas holdem poker hand, each player is dealt two cards from
a standard 52 card deck. How many possible two-card hands are there?
(b) At the start of a fve card draw poker hand, each player is dealt fve cards from
a standard 52 card deck. How many possible fve-card hands are there?
(c) What is the probability of being dealt a fush (5 cards all of the same suit) in
fve card draw? (In this question, any hand in which all 5 cards are the same
suit is considered to be a fush. In poker, straight fushes are considered to be
distinctly di˙erent from a fush, but they are a fush for the purposes of this
question.)

17. Suppose we are about to draw a simple random sample of 4 students from a class
of 200.

(a) How many di˙erent samples are possible?

(b) What is the probability that any one of the individual 200 students gets se-
lected in the sample?

18. (a) A committee of 5 people is about to be drawn from a group of 25 faculty. How
many di˙erent groups of 5 people can be chosen for the committee?
(b) A bussing company needs to assign bus drivers to 7 di˙erent routes. There are
30 bus drivers to choose from. How many di˙erent ways are there of assigning
7 bus drivers to the di˙erent routes?
Balka ISE 1.09 4.7. PROBABILITY AND THE LONG RUN 41

4.7 Probability and the Long Run

19. Tom and Pete are wasting time by repeatedly betting $1 on each toss of a fair coin.
If the coin comes up heads, Tom wins $1. If the coin comes up tails, Pete wins $1.
Which of the following statements are true?

(a) The proportion of times heads has come up will tend toward 0.5 as the number
of tosses tends to infnity.
(b) The amount of money the losing player owes the winning player will tend to
0 as the number of tosses tends to infnity.
(c) The probability they are exactly even after 1,000,000 tosses is greater than the
probability they are exactly even after 10 tosses.

4.8 Chapter Exercises

4.8.1 Basic Calculations

20. Suppose P (A) = 0.30, P (B) = 0.80, and P (A ∩ B) = 0.125.

(a) What is P (A ∪ B)?

(b) What is P ((A ∪ B)c )?
(c) What is P (A|B)?
(d) What is P (B|A)?
(e) Are A and B independent?
(f) What is P (Ac ∪ B)

21. Suppose A and B are two events such that P (A) = 0.5, P (B) = 0.2 and P (A∪B) =
0.68.

(a) What is P (A ∩ B)?

(b) What is P ((A ∪ B)c )?
(c) What is P (A|B)?
(d) What is P (B|A)?
(e) Are A and B independent?
(f) What is P (Ac ∪ B c )
(g) What is P (Ac ∩ B c )

22. Suppose P (A) = 0.2, P (B) = 0.2, P (A ∩ B) = 0.2. Which of the following
statements are true?

(a) A and B are mutually exclusive.

Balka ISE 1.09 4.8. CHAPTER EXERCISES 42

(b) A and B are independent.

(c) P (A|B) = 0.
(d) P (A|B) = 1.
(e) A and B c are not independent.
(f) P (A ∪ B) = P (A ∩ B).
(g) P (A|B) = P (B|A).

23. Suppose a random sample of 100 students who have taken both course A and course
B yielded the following:

Passed Course B Failed Course B

Passed Course A 71 19
Failed Course A 2 8

(a) If one of these students is randomly selected, what is the probability they
passed course A?
(b) If one of these students is randomly selected, what is the probability they
passed course A, if we know the student passed course B?
(c) If one of these students is randomly selected, are the events the student passed
course A and the student passed course B independent?

24. Suppose P (A) = 0.4, P (B) = 0.7, and the probability that A and B both occur is
0.2.

(a) What is P (A|(B ∪ A)c )?

(b) What is P (A ∩ B|A ∪ B)?

25. A balanced six-sided die is rolled once.

(a) What is the probability of rolling an even number on the top face?
(b) What is the probability of rolling a 3?
(c) What is the probability the number on the top face is a 3, given it is an even
number?
(d) What is the probability the number on the top face is at least 4, given it is an
even number?
(e) Are the events the number on the face is even and the number on the face is
more than 3 independent?
(f) What is the conditional probability the number on the face is 2 or less, given
the number is at least 4?
Balka ISE 1.09 4.8. CHAPTER EXERCISES 43

4.8.2 Concepts

26. Probabilities are often expressed in terms of odds. For example, we might say that
the odds in favour of an event are 2:7 or that the odds against an event are 6:1.
The odds in favour of an event is a ratio of two whole numbers that represents the
ratio of the probability of the event to the probability of its complement. The odds
against an event switches the order of the numbers (odds against represents the
ratio of the probability of the event’s complement to the probability of the event).
For example, suppose we are about to roll a fair die once and we are hoping to roll
a 2 (which has a probability of 16 ). The odds in favour of rolling a 2 are 1:5. (On
average, for every 2 that we roll, we will roll a di˙erent number 5 times.) The odds
against rolling a 2 are 5:1.

(a) Suppose we are about to roll a single die once. What are the odds in favour
of rolling a 5 or a 6? What are the odds against rolling a 5 or a 6?
(b) Suppose we are about to draw a single card from a well-shu˜ed deck. What
are the odds in favour of drawing a jack? What are the odds against drawing
a jack?
(c) If the odds in favour of an event are a:b, then on average, for every a occur-
rences of the event, the event will not happen b times, and thus the probability
of the event is a+b
a
. If the odds in favour of an event are 5:3, what is the prob-
ability of this event?
(d) If the odds against an event are 19:1, what is the probability of the event?

27. The following two identities are called De Morgan’s laws, and they frequently arise
in probability and logic. Verify these identities using Venn diagrams.

(a) (A ∪ B)c = Ac ∩ B c
(b) (A ∩ B)c = Ac ∪ B c

28. We have already been introduced to the addition rule for two events: P (A ∪ B) = P (A) + P (B) − P (A ∩ B)
Use an argument based on Venn diagrams to come up with the addition rule for
three events. That is, express P (A ∪ B ∪ C) in terms of the probabilities of A, B,
and C and the probabilities of their intersections.

29. Which one of the following statements is true? (Assume for the purposes of this
question that the events have non-zero probabilities of occurring.)

(a) Mutually exclusive events can be independent.

(b) Mutually exclusive events are always independent.
(c) Independent events are never mutually exclusive.
(d) Independent events are always mutually exclusive.

30. Which of the following statements are true? (There may be more than one true
Balka ISE 1.09 4.8. CHAPTER EXERCISES 44

statement; check all that are true.)

(a) If P (A) = 0.30 and P (B) = 0.30, then P (A ∩ B) = 0.09.

(b) If P (A) = 0.30 and P (B) = 0.30, and P (A ∪ B) = 0.60, then A and B are
independent.
(c) If P (A) = 0.50, and the probability that B occurs given A occurs is 0.50, then
P (A ∩ B) = 0.25.
(d) If P (A) = 0, then A is independent of any event.
(e) If P (B|A) = 0.80, then P (B) > 0.80.

31. Which of the following statements are true? (There may be more than one true
statement; check all that are true.)

(a) If P (A) = 0.6, P (B) = 0.7, then 0.3 ≤ P (A ∩ B) ≤ 0.6.

(b) If P (A) = 0.5, then for any event B: 0 ≤ P (A ∩ B) ≤ 0.5.
(c) If P (A|B) = 0, then A and B are mutually exclusive.
(d) Suppose that P (A) = 0.99. Event A and its complement are mutually exclu-
sive and independent.
(e) If P (A|B) 6= P (B) then A and B are not independent.

32. Two events A and B are mutually exclusive. If P (A) = 0.5 and P (B) = 0.3, are A
and B independent?

33. Suppose a fair coin is tossed twice, in such a fashion that the two tosses are inde-
pendent. Let A be the event that heads comes up on the frst toss, and B be the
event that heads comes up on the second toss. Which of the following 3 pairs of
events represent independent events.

(a) A and B
(b) A and B c
(c) A and Ac

34. Consider the diagrams shown in Figure 4.2. For each plot, assume that the area
of the square is 1, and that the area of each region is equal to its probability of
occurring. For which plots, if any, are A and B independent?

35. Suppose we randomly select a Canadian adult. Let A be the event that the person
is male. Let B be the event that the person weighs more than 100 kg. Which of
the following statements are true?

(a) P (A|B) < P (A).

(b) P (B|A) < P (B).
(c) A and B are mutually exclusive.
(d) A and B are not independent.
Balka ISE 1.09 4.8. CHAPTER EXERCISES 45

A∩ B A∩ Bc A∩ B A∩ Bc

Ac∩ B Ac∩ Bc Ac∩ B Ac∩ Bc

(a) (b)

A∩ B A∩ Bc A∩ B A∩ Bc

Ac∩ B Ac∩ Bc Ac∩ B Ac∩ Bc

(c) (d)

Figure 4.2: Venn diagrams for Question 34.

36. For the following pairs of events, state whether P (B|A) = P (B), P (B|A) > P (B),
or P (B|A) < P (B).

(a) For your current location:

A: It rains on Tuesday of next week.
B: It rains on Wednesday of next week.
(b) For your current location:
A: It rains on Tuesday of next week.
B: It rains on the day that is 10 years from today.
(c) For a randomly picked Canadian adult:
A: The person did not fnish high school.
B: The person has been to prison.

37. Two inspectors, A and B, work side by side at a factory. Their jobs involve in-
specting the quality of each part that is produced and giving each part a pass or
fail rating. They both just started the job, and are both very lazy. They decide
to randomly choose between passing and failing each part. For each part, one of
them tosses a coin twice and they both make their decisions based on these 2 tosses.
Inspector A passes the part if the frst coin comes up heads and fails it otherwise.
Inspector B passes the part if heads come up exactly once and fails it otherwise.
Consider the events:

A: Inspector A passes the part. B: Inspector B passes the part.

Balka ISE 1.09 4.8. CHAPTER EXERCISES 46

(a) Are A and B mutually exclusive?

(b) Are A and B independent?

38. Daniel has to do an online assignment for one of his classes. The assignment consists
of 20 multiple choice questions, where there are 5 answer options for each question.
Daniel knows nothing about the material, and if he were to do the assignment on
his own he would guess randomly on each question. Daniel’s friend Magnus has an
80% chance of getting a passing grade on the assignment. (Magnus works a little
harder than Daniel.) Daniel is tempted to cheat on the assignment by copying his
friend Magnus’s answers, but Daniel doesn’t like cheating, and he knows that the
university considers cheating on an assessment to be a serious case of academic
misconduct. Daniel is torn between cheating and doing the assignment on his own,
so he elects to let the fates decide. He will toss a coin once, and if it comes up
heads he will copy every answer o˙ his friend Magnus. If the coin comes up tails,
Daniel will do the assignment himself, guessing at every answer.

(a) Estimate the probability that Daniel passes the assignment. (We haven’t
learned how to calculate the exact probability of this yet, but using a sim-
ple argument we should be able to come pretty close.)
(b) Are the events Daniel gets a passing grade on the assignment and Magnus gets
a passing grade on the assignment mutually exclusive?
(c) Are the events Daniel gets a passing grade on the assignment and Magnus gets
a passing grade on the assignment independent?

39. Test your conceptual understanding: Which of the following statements are true?
You should be able to explain why a statement is true or why a statement is false.

(a) Probabilities can be negative.

(b) For any two events A and B, P (A|B) > P (A).
(c) If A and B are independent, then P (A|B) = 0.
(d) If two cards are drawn from a standard deck, the probability both cards are
red is greater if the cards are drawn with replacement than if they were drawn
without replacement.
(e) If two events A and B are independent, then their complements (Ac and B c )
are also independent.
(f) If A and B are mutually exclusive, then P (A|B) = 0.

40. Test your conceptual understanding: Which of the following statements are true?
You should be able to explain why a statement is true or why a statement is false.

(a) If P (A) = P (B), then A and B are independent.

(b) If P (A) > P (B), then P (A|B) ≥ P (B|A).
(c) If A and B are independent, then P (A ∪ B) = P (A) + P (B).
Balka ISE 1.09 4.8. CHAPTER EXERCISES 47

(d) If A and B are mutually exclusive, then P (A ∪ B) = P (A) + P (B).

(e) If P (A|B) = 1, then P (B|A) = 1.

4.8.3 Applications

41. In a 2008 study that was picked up by major news networks, two Manhattan
teenagers investigated whether restaurants and grocery stores that served sushi
were mislabelling the fsh that was used. They sent samples to the University of
Guelph for analysis, and found that 25% of the fsh samples that could be geneti-
cally identifed were mislabelled. Often, cheaper fsh was served instead of a more
expensive variety. Assume for the purposes of this question that a randomly se-
lected sushi dish (that contains fsh) at a randomly selected restaurant has a 25%
chance of having mislabelled fsh.

(a) Suppose you go to a randomly selected sushi restaurant. You order 4 randomly
selected types of sushi (the fsh variety). Assuming independence, what is the
probability that you get at least one piece that is mislabelled?
(b) True or False? The independence assumption in the previous question is per-
fectly justifed.

42. Suppose a group of 50 students is made up of 20 males and 30 females. Twenty fve
of these students have part time jobs. 15 of the males do not have a job. Suppose
one of these 50 students is randomly selected.

(a) What is the probability they do not have a job?

(b) What is the probability they are a male that does not have a job?
(c) If they have a part-time job, what is the probability they are male?
(d) If they are a female with a job, what is the probability they are male?
(e) For these 50 students, is job status independent of gender?

43. Is there a relationship between education and smoking status among French men? A
study2 measured several variables on 459 healthy men in France who were attending
a clinic for a check up. Table 4.2 gives the results of the study. Suppose one of
these 459 men is randomly selected.

(a) What is the probability the man is a smoker?

(b) What is the probability the man has a university education?
(c) Given the man is a smoker (moderate or heavy), what is the probability that
he has a university education?
2
Marangan et al. (1998). Diet, antioxidant status, and smoking habits in French men. American Journal
of Clinical Nutrition, 67:231–239.
Balka ISE 1.09 4.8. CHAPTER EXERCISES 48

Nonsmoker Ex-smoker Moderate Heavy

Primary School 56 54 41 36
Secondary School 37 43 27 32
University 53 28 36 16

Table 4.2: Smoking status and level of education for 459 French men.

(d) Given the man is a smoker (moderate or heavy), what is the probability that
he has at least a high school education?
(e) Given the man is a nonsmoker, what is the probability that he has less than
a high school education?
(f) Given the man has only a secondary school education, what is the probability
that he is a smoker?
(g) Given the man has at least a secondary school education, what is the proba-
bility that he is a smoker?
(h) Are they events the man has a university education and the man is a smoker
independent?
(i) Are they events the man has a university education and the man is a smoker
mutually exclusive?

44. Is there a relationship between fatty fsh consumption and the rate of prostate
cancer? A study3 followed 6272 Swedish men for 30 years. They were categorized
according to their fsh consumption, and to whether they developed prostate cancer.
The following table summarizes the results. Suppose one of these 6272 men is
randomly selected.

Fish consumption
Never/Seldom Small Moderate Large
Prostate cancer 14 201 209 42
No prostate cancer 110 2420 2769 507

(a) What is the probability their fsh consumption was large?

(b) What is the probability their fsh consumption was small, moderate, or large?
(c) What is the probability they got prostate cancer?
(d) What is the probability they got prostate cancer, given their fsh consumption
was never/seldom?
(e) What is the probability they got prostate cancer, given their fsh consumption
was large?
(f) What is the probability their fsh consumption was large, given they got
prostate cancer?
(g) What is the probability they got prostate cancer, given their fsh consumption
3
Terry et al. (2001). Fatty fsh consumption and risk of prostate cancer. Lancet, 357:1764–1766.
Balka ISE 1.09 4.8. CHAPTER EXERCISES 49

was small or moderate?

45. The breakdown of male and female full-time undergraduate students at the Univer-
sity of Guelph, University of Waterloo, and Wilfrid Laurier University in the fall
semester of 2009 is given in Table 4.3.

University Male Female Total

Guelph 6,366 10,417 16,783
Waterloo 13,770 10,607 24,377
Laurier 4,803 6,254 11,057
Total 24,939 27,278 52,217

Table 4.3: Undergraduate enrolment in the fall semester of 2009.

If one of these 52,217 students is randomly selected:

(a) What is the probability they attended the University of Guelph that semester?
(b) What is the probability they are female?
(c) What is the probability they attended the University of Guelph, given they
are female?
(d) What is the probability they are female, given they attended the University
of Guelph?
(e) What is P (F |G∪Lc ), where F represents the event that the randomly selected
student is female, G represents the event that the randomly selected student
attended the University of Guelph that semester, and L represents the event
that the randomly selected student attended Wilfrid Laurier University that
semester?

46. A restaurant conducts a survey of customers as they are leaving. The following
table gives a summary of the results.

Dining experience was good Dining experience was not good

Men 23 7
Women 17 18

(a) If one of these 65 customers is selected at random, what is the conditional

probability the dining experience was good, given they are male.
(b) If one of these 65 customers is selected at random, what is the conditional
probability they are a man, given they are a woman.
(c) If one of these 65 customers is randomly selected, are the events the customer
is a man and the dining experience was good independent?
(d) Are the events the customer is a man and the customer is a woman indepen-
dent? Are they mutually exclusive?
Balka ISE 1.09 4.8. CHAPTER EXERCISES 50

47. A student is worried about sleeping in and missing their fnal exam, so they set two
battery powered alarm clocks. The probability the frst alarm will go o˙ is 0.99,
and the probability the second alarm will go o˙ is 0.98. Assume that these clocks
work independently. (Independence might be a sketchy assumption in this spot,
but let’s go with it.)

(a) What is the probability that both alarms go o˙?

(b) What is the probability that neither alarm will go o˙?
(c) What is the probability that at least one alarm goes o˙?
(d) Which one of the above is the most important question, given the nature of
the problem?

48. Two cards are drawn with replacement from a standard deck.

(a) What is the probability that both cards are kings?

(b) Given there is at least one king drawn, what is the probability both cards are
kings?
(c) What is the probability that exactly one card is red?
(d) What is the probability that at least one card is red?
(e) What is the probability at least one card is a king, and at least one card is
red?

49. Two cards are drawn without replacement from a standard deck.

(a) What is the probability that both cards are kings?

(b) Given there is at least one king drawn, what is the probability both cards are
kings?
(c) What is the probability that exactly one card is red?
(d) What is the probability that at least one card is red?
(e) What is the probability at least one card is a king, and at least one card is
red?

50. Mateo is both rich and a horrible driver. He is going to drive to a convenience store
to buy cigarettes, and he must decide which one of his two cars he will take. He
decides to fip a fair coin twice, and if it comes up heads both times he will take his
Ferrari, and otherwise he will take his Lexus. If he takes his Ferrari he has a 0.08
probability of getting into an accident. If he takes his Lexus, the probability that
he gets into an accident is 0.04.

(a) What is the probability that Mateo gets into an accident?

(b) Given he takes his Ferrari, what is the probability that he gets into an accident?
(c) Given he gets into an accident, what is the probability he took the Ferrari?

51. At a certain gas station, 80% of the people buy regular gas, 5% buy midgrade, and
Balka ISE 1.09 4.8. CHAPTER EXERCISES 51

15% buy premium. Suppose if a randomly selected person pumps regular gas, there
is a probability of 0.02 that they leave without paying. If they pump midgrade, the
probability is 0.04, and if they pump premium the probability is 0.14.

(a) For a randomly selected gas pumper, what is the probability they leave without
paying?
(b) If they leave without paying, what is the probability that they pumped pre-
mium gas?

52. You need to buy a part from a supplier. The supplier has 3 di˙erent sources for
this part. The supplier will get the part from Source A with probability 0.55, from
Source B with probability 0.35, and from Source C with probability 0.10. If the
part comes from Source A, there is a 40% chance it will last for a year. If the part
comes from Source B, there is an 90% chance it will last for a year. If the part
comes from Source C, there is a 55% chance it will last for a year. You order a part
from the supplier.

(a) What is the probability the part lasts for a year?

(b) Given the part lasts for a year, what is the probability that it came from
Source A?

53. In sample surveys with sensitive questions, respondents are likely to lie, and thus
the survey results would not refect reality. (For example, it may be diÿcult to get a
truthful answer to the question, “have you ever fantasized your spouse was dead?”)
One way to combat this e˙ect is to implement a randomized response survey.
For example, in a telephone survey an individual may be told to toss a coin before
responding. If the coin comes up heads, they should respond “Yes”, regardless of the
truth. If coin comes up tails they should tell the truth. Incorporating a randomized
response adds a level of variability to the responses, which is undesirable. But
respondents may be more willing to answer truthfully, since the person on the
other end of the telephone will not know if the respondent is telling the truth, or if
they responded that way simply because the coin fell a certain way.

(a) Suppose that in reality, 20% of a population has stolen from their employer.
What is the probability a randomly selected person from this population re-
sponds “Yes” to the question “Have you stolen from your employer?” (Assuming
the person follows the directions given above.)
(b) Assume again that 20% of a population has stolen from their employer. If a
randomly selected person responds “Yes”, what is the probability they have
stolen from their employer?
(c) Suppose the probability that a person has stolen from their employer is not
known (which is typically the case). If 67% of the respondents answer “Yes”
to the question, what is the best estimate of the proportion of individuals who
have stolen from their employer?
Balka ISE 1.09 4.8. CHAPTER EXERCISES 52

54. Participants in a survey were asked to use a randomized response to the question
“have you ever cheated on your spouse?” They were asked to roll a die, and if the die
came up with a 5 or a 6 they were to tell the truth. If the die came up with a 1, 2,
3, or 4 they were to answer “yes”, regardless of the truth. If 82% of the respondents
answer “yes” to the question, what is the best estimate of the proportion of these
people who have actually cheated on their spouse?

55. In a certain population of mice, each mouse has a one-in-a-thousand chance of

having a specifc genetic disorder. You intend to sample mice and check to see if
they have this disorder. Suppose that the test for this disorder works perfectly, and
the mice can be considered independent.

(a) If you sample 30 mice, what is the probability that none of them have the
genetic disorder?
(b) What is the probability that more than 500 mice will be sampled before fnding
the frst mouse with the disorder?
(c) How many mice would need to be sampled before there is at least a 50% chance
of fnding at least one with the disorder?

56. You have an expensive system that relies on an important component. This com-
ponent has a probability of failure of 0.32. You feel this probability is much too
high, and you have designed a system in which you can install as many independent
components as you like. The system connects these devices in parallel, so if at least
one of them works the system works.

(a) If you connect 4 of these components in parallel, what is the probability of

failure of the system?
(b) What is the smallest number of these components that you would have to
connect in parallel for the probability of failure to be less than 0.0001?

57. An important satellite guidance system relies on a very fragile component that has a
0.77 probability of failure. Since this failure probability is very high, designers want
to put in independent back-up components (each of these also has a probability of
failure of 0.77). They wish to connect them in parallel such that the guidance
system works as long as at least one of the components works.

(a) If you connect 5 of these components in parallel, what is the probability the
system fails?
(b) If you connect 10 of these components in parallel, what is the probability the
system fails?
(c) What is the smallest number of components that must be connected in order
for the probability the system works to be at least 0.99990?

58. You intend to do research into people with a certain rare genetic marker. The
Balka ISE 1.09 4.8. CHAPTER EXERCISES 53

probability a randomly selected person has the marker is 0.0001. You need people
with the marker for your experiment. Assume that you can randomly sample people
from the population.

(a) What is the probability the frst person with the marker occurs on the 2nd
person sampled?
(b) What is the probability the frst person with the marker occurs on the 10th
person sampled?
(c) What is the probability that the frst person with the marker occurs on or
before the 2,000th person sampled?

59. A large art museum has two painting experts on sta˙. One of their jobs is to
assess whether given artwork is authentic (they decide whether it is genuine or a
forgery). Expert A and Expert B both look at each artwork individually, and arrive
at their own conclusions. A summary of the results of inspections of 200 pieces of
art follows.
Judged genuine by B Judged a forgery by B
Judged genuine by A 174 4
Judged a forgery by A 2 20

(a) If one of these pieces of art is randomly selected, what is the probability that
Expert A thought it to be genuine?
(b) One of these pieces of art is randomly selected. Given Expert B thought it a
forgery, what is the probability that Expert A also thought it a forgery?
(c) Are events Expert A thinks the piece is a forgery and Expert B thinks the piece
is a forgery mutually exclusive? Are they independent?

60. John has a crush on Stephanie. A friend of John's is having a party on Friday
night. If Stephanie goes to the party, the probability that John will also go is 0.94.
If Stephanie does not go to the party, the probability that John goes is 0.03.

(a) If the probability that Stephanie does not go to the party is 0.20, what is the
probability that John goes?
(b) Suppose now that the probability Stephanie goes to the party is unknown, but
the conditional probabilities that John goes (0.94, 0.03) are still the same. If
the probability that John goes to the party is 0.38, what is the probability
that Stephanie goes to the party?
(c) Are the events John goes to the party and Stephanie goes to the party mutually
exclusive? Are they independent?

61. A certain genetic defect a˙ects 0.001% of the population. A test is available for
this defect. If someone has the defect, the test will fail to detect it with probability
0.01. If somebody does not have the defect, the test will give a false positive with
Balka ISE 1.09 4.8. CHAPTER EXERCISES 54

probability 0.005.

(a) What is the probability that a randomly selected person tests positive?
(b) Given a randomly selected person tests positive for this defect, what is the
probability they have the defect?

62. The following table is based on a Statistics Canada publication.4

40 years 0.99864
41 years 0.99852
42 years 0.99837
43 years 0.99818
44 years 0.99798
45 years 0.99775
46 years 0.99751
47 years 0.99726
48 years 0.99702
49 years 0.99678

The table represents part of a life table, which gives various estimated probabilities
of survival. The probabilities given here are estimated conditional probabilities. For
example, the estimated probability that an exactly 40 year old Ontario man survives
until his 41st birthday is .99864. The estimated probability that an exactly 41 year
old Ontario man survives until his 42nd birthday is .99852, etc.

(a) Based on this table, what is the probability an Ontario man, alive on his 47th
birthday, survives until his 48th?
(b) Based on this table, what is the probability an Ontario man, alive on his 47th
birthday, dies before his 48th?
(c) Based on this table, what is the probability a male child born in Ontario lives
until their 40th birthday?
(d) Based on this table, what is the estimated probability that an Ontario man
who has survived until his 45th birthday dies when he is 48? That is, he reaches
his 48th birthday but dies before his 49th .
(e) What is the estimated probability that a randomly selected Ontario man, alive
on his 45th birthday, survives until his 48th?

4
The full table is available at http://www.statcan.gc.ca/pub/84-537-x/t/txttables/ontm.txt. Ac-
cessed January 11 2014.
Chapter 5

Discrete Random Variables and

Discrete Probability Distributions

J.B.’s strongly suggested exercises: 2, 5, 6, 7, 8, 10, 13, 14, 15, 19, 20, 22, 23, 25, 26, 28,
30, 48, 51, 53, 54, 57, 58, 62, 64, 66 (a-d).

5.1 Introduction

5.2 Discrete and Continuous Random Variables

1. For each of the following random variables, state whether they are discrete or con-
tinuous.

(a) The number of poker hands dealt in a casino in an hour.

(b) The time until completion of a randomly selected poker hand in a casino.
(c) The amount of cheese on a randomly selected cheeseburger at a fast food
restaurant.
(d) The number of moves in a game at the World Chess Championship.
(e) The duration of a match at the World Chess Championship.
(f) The number of die rolls required to roll a six 18 times in a row.
(g) The sum of the numbers that come up on the top face when a die is rolled 400
times, divided by 3617.

55
Balka ISE 1.09 5.3. DISCRETE PROBABILITY DISTRIBUTIONS 56

5.3 Discrete Probability Distributions

2. Suppose that two cards are drawn without replacement from a standard 52 card
deck. Let X represent the number of hearts drawn. Find the probability distribu-
tion of X.

3. Consider the following probability distribution of a random variable X.

x 10 20 30 40
p(x) 0.2 0.2 0.5 ?

(a) Is X discrete or continuous?

(b) What is the value of the missing probability?
(c) What is the most likely value of X?
(d) What is P (X < 32)?
(e) What is the conditional probability X is less than 25, given X is less than 35?

4. Which of the following are valid discrete probability distributions?

x −10 1.7 30000

(a)
p(x) 0.2 0.2 0.6

x 1 2 3
(b)
p(x) 0.2 0.2 0.2

x 0.1 0.2 0.3 4000 500000

(c)
p(x) 0.2 0.2 0.2 0.2 0.2

x 1 2 3 4 5
(d)
p(x) −0.2 0.3 0.3 0.4 0.2

5.3.1 The Expectation and Variance of Discrete Random Variables

5.3.1.1 Calculating the expected value and variance of a discrete ran-

dom variable

5. Consider the following probability distribution of a random variable X.

x −2.5 0 2.5 5.0 7.5

p(x) 0.10 0.10 0.10 0.40 ?
Balka ISE 1.09 5.3. DISCRETE PROBABILITY DISTRIBUTIONS 57

(a) Is X discrete or continuous?

(b) What is the value of the missing probability?
(c) What is the most likely value of X?
(d) What is P (−1 < X < 4)?
(e) What is the expected value of X?
(f) What is the standard deviation of X?

6. Hellin’s law is a rough guideline that gives approximate probabilities of multiple

births in pregnancies that are not the result of fertility treatments. Hellin’s law
states that approximately 1 in 89 pregnancies result in twins, approximately 1
in 892 pregnancies result in triplets, approximately 1 in 893 pregnancies result in
quadruplets, and so on.

The following table gives the approximate distribution of the number of babies
delivered at birth for pregnancies that are not the result of fertility treatments.
(The values in the table are roughly based on Hellin’s law. Quintuplets and higher
order multiple births are extremely rare and are ignored for the purposes of this
question.)

Number of babies 1 2 3 4
Probability 0.988 0.01187 0.000128 0.000002

For the following questions, “pregnancy” refers to a pregnancy that is not the result
of fertility treatments.

(a) What is the mean number of children per pregnancy?

(b) What is the standard deviation of the number of children per pregnancy?
(c) What is the probability of a multiple birth?
(d) Given the pregnancy results in a multiple birth, what is the probability it
results in twins?
(e) Given the pregnancy results in triplets or quadruplets, what is the probability
it results in triplets?

7. Which of the following statements are true?

(a) Probabilities of impossible events can be negative.

(b) If a random variable can take on an infnite number of possible values, then it
cannot be a discrete random variable.
(c) The mean of a discrete random variable cannot be negative.
(d) The standard deviation of a discrete random variable cannot be negative.
(e) The mean of a discrete random variable cannot be greater than its standard
deviation.
Balka ISE 1.09 5.4. THE BERNOULLI DISTRIBUTION 58

5.3.1.2 Properties of Expectation and Variance

8. Suppose:

• X is a random variable with µX = 8.7 and σX = 18.5.

• Y is a random variable with µY = 14.9 and σY = 21.0.
• X and Y are independent.

(a) What is the mean of X + Y ?

(b) What is the standard deviation of X + Y ?
(c) What is the standard deviation of X − Y ?

9. Twenty university students are taking both Course A and Course B. In course A the
professor is very conscientious, and she makes an e˙ort to make the tests consistent
with and representative of the material covered in the class. Course B is taught by
a professor who doesn’t give any tests, and at the end of the semester he simply
randomly assigns grades to students. Is a student’s grade in course A independent
of their grade in course B? Explain.

10. An assembly procedure consists of three independent steps. Step 1 has a mean time
to completion of 6 minutes and a standard deviation of 1.2 minutes. Step 2 has a
mean time to completion of 125 minutes and a standard deviation of 15.0 minutes.
Step 3 has a mean time to completion of 17 minutes and a standard deviation of 3
minutes.

(a) What is the mean time to completion of the entire procedure?

(b) What is the standard deviation of the time to completion of the entire proce-
dure?
(c) Suppose we wish to measure time in hours instead of minutes. What are the
mean and variance of the time (in hours) to completion?

5.4 The Bernoulli Distribution

11. Suppose a single card is drawn from a well-shu˜ed 52 standard deck.

(a) What is the distribution of the number of eights that are drawn?
(b) What is the mean number of eights that are drawn?
(c) What is the standard deviation of the number of eights that are drawn?
Balka ISE 1.09 5.5. THE BINOMIAL DISTRIBUTION 59

5.5 The Binomial Distribution

12. Suppose X is a binomial random variable with parameters n = 15 and p = 0.2.

(a) What is P (X = 3)?

(b) What is P (X ≤ 3)?
(c) What is P (X > 3)?
(d) What is P (2 ≤ X < 5)?
(e) What is P (X = 3|X ≤ 4)
(f) What is the mean of X?
(g) What is the standard deviation of X?

13. According to the United States Centers for Disease Control and Prevention, approx-
imately 3% of babies in the United States are born with major structural or genetic
birth defects. Suppose 50 newborns in the United States are randomly selected.

(a) What is the probability that exactly 2 have major structural or genetic birth
defects?
(b) What is the probability that no more than 2 have major structural or genetic
birth defects?
(c) What is the probability that exactly 10 have major structural or genetic birth
defects?
(d) What is the expectation of the number with major structural or genetic birth
defects?
(e) What is the standard deviation of the number with major structural or genetic
birth defects?

14. Which of the following statements are true?

(a) The variance of a binomial distribution is always greater than the mean.
(b) A binomial random variable can take on negative values.
(c) The mean of a binomial random variable can be negative.
(d) The variance of a binomial random variable can be negative.
(e) A binomial random variable is a discrete random variable.
(f) Every discrete random variable is a binomial random variable.
(g) A binomial random variable represents a count.
(h) The mean of a binomial random variable is np.
(i) The standard deviation of a binomial random variable is np(1 − p).
p

(j) For any given n, the variance of a binomial random variable is greatest when
p = 0.5.
Balka ISE 1.09 5.5. THE BINOMIAL DISTRIBUTION 60

5.5.1 Binomial or Not?

15. Which of the following random variables have a binomial distribution?

(a) The number of hearts if 10 cards are drawn without replacement from a stan-
dard deck.
(b) The number of times heads comes up if a biased coin is tossed 10 times.
(c) The number of putts a golfer holes in their next 10 attempts.
(d) The number of newborn babies in a maternity ward on a randomly selected
day.
(e) The amount of money withdrawn by a randomly selected customer at a bank.

16. Which of the following random variables have a binomial distribution?

(a) The number of hearts if 10 cards are drawn with replacement from a standard
deck.
(b) The number of fngers on a randomly selected newborn.
(c) The number of times a sum of 7 is rolled when a pair of dice is rolled 14 times.
(d) The number of children a randomly selected woman has.
(e) The number of times a randomly selected person uses the washroom on a
randomly selected day.
(f) The weight of 15 randomly selected playing cards.

5.5.2 A Binomial Example with Probability Calculations

17. A friend of yours says they have extra-sensory perception (ESP), and claims they
can pull the king of spades from a well shu˜ed 52-card deck approximately 20% of
the time. You think they are being foolish and you decide to test their claim. You
open a brand new deck of cards, shu˜e it well, and ask them to draw the king of
spades. You repeat this procedure until they have drawn a card 50 times. They
manage to draw the king of spades twice in the 50 draws.

(a) If your friend does not have ESP, and is merely randomly pulling a card each
time, what is the probability they draw the king of spades at least twice in 50
draws?
(b) Does this experiment give strong evidence that your friend does not have ESP?
Balka ISE 1.09 5.6. THE HYPERGEOMETRIC DISTRIBUTION 61

5.6 The Hypergeometric Distribution

18. An urn contains 22 white and 18 red balls. Six balls are randomly chosen without
replacement.

(a) What is the probability that exactly 4 white balls are chosen?
(b) What is the probability that no more than 1 white ball is chosen?
(c) What is the mean number of white balls chosen? What is the mean number
of red balls chosen?

19. You have an idea to start a business that delivers pizzas using stretch limousines.
You need to hire 5 drivers, and 20 people apply for the positions. Unknown to you,
7 of these 20 people have criminal records. Suppose you (foolishly) decide to hire a
random sample of 5 drivers from the 20 applicants.

(a) What is the probability that exactly three of those hired have a criminal
record?
(b) What is the probability that exactly two of those hired have a criminal record?
(c) What is the probability that at least one of those hired has a criminal record?
(d) On average, how many of the 5 people you hire will have a criminal record?

20. Is it reasonable to use the binomial distribution to approximate the hypergeometric

distribution in the following scenarios? (It depends, of course, on how accurate we
need to be, and in most situations we prefer the exact answer to an approximation.
Answer these questions based on the rough guideline given in the text.)

(a) Suppose that 8 people are randomly selected from a room containing 10 men
and 10 women. Is it reasonable to use the binomial distribution to approximate
the probability that exactly 4 of the people selected are men?
(b) Twelve balls are drawn without replacement from a large urn containing 100
white balls and 200 red balls. Is it reasonable to use the binomial distribution
to approximate the probability that exactly 3 balls are white?
(c) A card is drawn from a well-shu˜ed deck. The card is placed back in the deck,
the cards are shu˜ed and another card is drawn. The process is repeated for a
total of 60 draws. Is it reasonable to use the binomial distribution to calculate
the probability that exactly 7 cards are kings?
Balka ISE 1.09 5.7. THE POISSON DISTRIBUTION 62

5.7 The Poisson Distribution

5.7.1 Introduction

21. Suppose X is a Poisson random variable with parameter λ = 5.

(a) What is P (X = 2)?

(b) What is P (X ≤ 2)?
(c) What is P (X > 2)?
(d) What is P (2 < X ≤ 5)?
(e) What is P (X = 2|X ≤ 3)
(f) What is the mean of X?
(g) What is the standard deviation of X?

22. On a certain region of the Florida coast, shark attacks on humans occur at a rate
of approximately 4 attacks per year. Assume that the number of attacks follows a
Poisson distribution.

(a) In a given year, what is the probability there is exactly one attack?
(b) In a given year, what is the probability there are no more than 2 attacks?
(c) What is the probability that in a given two year period there are exactly 4
attacks?

23. Which of the following statements are true?

(a) A Poisson random variable can take on a countably infnite number of possible
values.
(b) A Poisson random variable can take on negative values.
(c) A Poisson random variable represents a count of the number of occurrences of
an event.
(d) The mean and variance of a Poisson random variable are always equal.
(e) If X has a Poisson distribution, then P (X = 0) < P (X = 1).
(f) Every random variable that represents a count is a Poisson random variable.

5.7.2 The Relationship Between the Poisson and Binomial Distri-

butions

24. In which one of the following situations would the Poisson distribution provide the
most reasonable approximation to the binomial distribution?

(a) X has a binomial distribution with n = 5 and p = 0.5.

Balka ISE 1.09 5.8. THE GEOMETRIC DISTRIBUTION 63

(b) X has a binomial distribution with n = 10, 000 and p = 0.9.

(c) X has a binomial distribution with n = 10 and p = 0.2.
(d) X has a binomial distribution with n = 500 and p = 0.01.
(e) X has a binomial distribution with n = 5 and p = 0.001.

25. Protanopia is a kind of colour blindness that a˙ects approximately 1% of males

(and a much smaller percentage of females). Individuals with protanopia do not
perceive red light normally, and have diÿculty distinguishing between red and green
and between red and blue. Suppose 100 males are randomly selected from a large
population in which 1% of the males have protanopia.

(a) Using the binomial probability mass function, fnd the probability that exactly
2 of these 100 males have protanopia. Also fnd the approximate probability
based on the Poisson distribution.
(b) Using the binomial probability mass function, fnd the probability that no
more than 2 of these 100 males have protanopia. Also fnd the approximate
probability based on the Poisson distribution.
(c) Based on the binomial distribution, what is the variance of the number of males
with protanopia? What is the variance based on the Poisson approximation?

5.7.3 Poisson or Not? More Discussion on When a Random Vari-

able has a Poisson distribution

26. Suppose it is known that the number of students entering a University Centre
averages 4.5 per minute between noon and 12:30 pm. Would the number of students
that enter this University Centre in a randomly selected minute in this time frame
follows a Poisson distribution with λ = 4.5? Why or why not?

27. Suppose a very large pasture has a large number of cows, with an average rate of 1
cow per 100 m2 . Would the number of cows in a randomly selected 100 m2 area of
this pasture follow a Poisson distribution?

5.8 The Geometric Distribution

28. Suppose a patient at a hospital in Canada is in dire need of a plasma transfusion.

The person has type AB blood, and people with type AB blood can receive plasma
only from other individuals with type AB blood. The hospital is out of AB plasma,
but they have a long list of possible donors of unknown blood type. The donors are
Canadian, and approximately 3% of the Canadian population has type AB blood.
Balka ISE 1.09 5.8. THE GEOMETRIC DISTRIBUTION 64

Suppose these potential donors can be thought of as a random sample from the
Canadian population.

(a) If potential donors are tested for the AB blood type, what is the probability
the frst donor with type AB blood occurs on the ffth person tested?
(b) If potential donors are tested for the AB blood type, what is the probability
the frst donor with type AB blood occurs on or before the third person tested?
(c) If potential donors are tested for the AB blood type, what is the probability
the frst donor with type AB blood occurs after the 30th person tested?
(d) What is the mean number of potential donors that must be tested in order to
fnd one with type AB blood?
(e) What is the standard deviation of the number of potential donors that must
be tested in order to fnd one with type AB blood?

29. Tim Horton’s co˙ee shop chain has long held a “Roll up the Rim” contest, where
purchasers of a hot beverage can roll up the rim of the cup, possibly revealing that
the cup is a prize winner. The probability of winning a prize on each cup changes
slightly from year to year, but it is often approximately 16 . Suppose that while the
contest is running, you often buy cups of co˙ee from Tim Horton’s.

(a) What is the probability your frst winning cup comes on the fourth cup that
you purchase? (While the probability of winning changes a tiny amount from
cup to cup, as cups are revealed as winners or non-winners, there are millions
of cups so the changes in probability are minuscule. For the purposes of these
questions, assume that the cups can be considered independent.)
(b) What is the probability your frst win comes on or before the fourth cup?
(c) What is the probability your frst win comes after the 20th cup?
(d) On average, how many cups will need to be purchased in order to get the frst
winner?

30. Which of the following statements are true?

(a) The mean of the geometric distribution is p1 .

(b) The most likely value of a geometric random variable is always 1.
(c) If p > 0.5, then P (X = 2) < P (X = 3).
(d) The variance of the geometric distribution is always 1.
(e) A geometric random variable can take on a countably infnite number of pos-
sible values.
(f) If X has a geometric distribution, then Y = X1 also has a geometric distribu-
tion.
Balka ISE 1.09 5.9. THE NEGATIVE BINOMIAL DISTRIBUTION 65

5.9 The Negative Binomial Distribution

31. In his younger days, your author frequently practiced his golf game. While practic-
ing around the greens, he would often play a game in which he had to hole a certain
number of chips before allowing himself to leave. Suppose on a certain afternoon,
he decided that he would allow himself to leave once he chipped-in 4 times. (He
must hole out on a chip 4 times before leaving.) Suppose that his probability of
chipping-in on any individual chip is approximately 0.05, and that the chips can be
considered independent. (While the probability of chipping-in would vary a little
from chip to chip, and the chips would not truly be independent, these assumptions
provide a reasonable approximate model.)

(a) What is the probability that the fourth chip-in occurs on the 20th attempt?
(b) What is the probability that the fourth chip-in occurs on the 20th or 21st
attempt?
(c) What is the mean number of attempts required to chip-in 4 times?
(d) What is the standard deviation of the number of attempts required to chip-in
4 times?
(e) Challenge: What is the probability that it takes more than 100 attempts for
him to chip-in 4 times?

32. Suppose a door-to-door salesman needs to make 3 more sales in order to reach his
monthly sales goal. He knows from a large body of past experience that any time
he knocks on a door he has approximately a 1.5% chance of making a sale. Assume
that this is the correct probability of making a sale each time he knocks on a door,
and that it is reasonable to assume independence between doors.

(a) What is the probability that his 3rd sale occurs on the 50th door?
(b) What is the probability that his 3rd sale occurs on the 200th door?
(c) What is the mean number of doors he must approach in order to make his 3
sales?
(d) What is the standard deviation of the number of doors he must approach in
order to make his 3 sales?
(e) Challenge: What is the probability he must knock on more than 500 doors in
order to make his 3 sales?

5.10 The Multinomial Distribution

33. There are 3 main types of fngerprint pattern: arches, loops, and whorls. In fn-
gerprint analysis and classifcation, each fnger falls into one of these 3 categories.
Balka ISE 1.09 5.10. THE MULTINOMIAL DISTRIBUTION 66

(In more complicated classifcation systems there are various subcategories of these
groups.) Suppose that the distribution of fngerprint pattern on the left thumb of
adults in Libya follows the following distribution:1

Pattern Arches Loops Whorls

Probability 0.12 0.49 0.39

Suppose 10 Libyan adults are randomly selected and the fngerprint pattern on the
left thumb is identifed.

(a) What is the probability that 2 have arches, 3 have loops, and 5 have whorls?
(b) What is the probability that 4 have arches, 4 have loops, and 2 have whorls?
(c) What is the probability that exactly 1 has arches?
(d) What is the expectation of the number that have arches?

34. A typical European roulette wheel has 18 red slots, 18 black slots, and 1 green
slot. The wheel is spun and a ball lands in one of these equally likely slots. The
distribution of the colour of the slot the ball lands in is:

Colour Red Black Green

Probability 18
37
18
37
1
37

(a) In the next 20 spins of the wheel, what is the probability the ball lands in a
red slot 8 times, a black slot 10 times, and the green slot 2 times?
(b) In the next 20 spins of the wheel, what is the probability the ball lands in a
red slot 14 times, a black slot 3 times, and the green slot 3 times?
(c) In the next 20 spins of the wheel, what is the probability the ball lands in a
red slot exactly 11 times?
(d) What is the mean number of times the ball lands in a red slot in the next 20
spins?

35. Suppose X1 , X2 , . . . , Xk have a multinomial distribution with parameters n and

p1 , p2 , . . . , pk . Which of the following statements are true?

(a) X2 has a binomial distribution with parameters n and p2 .

(b) E(Xi ) = npi .
(c) V ar(Xi ) = npi (1 − pi ).
Pk
(d) i=1 pi = 1.
1
Based on a study by Fayrouz et al. (2012). Relation between fngerprints and di˙erent blood groups.
Journal of Forensic and Legal Medicine, 19:18–21
Balka ISE 1.09 5.11. CHAPTER EXERCISES 67

5.11 Chapter Exercises

5.11.1 Basic Calculations

36. Consider the following discrete probability distribution:

x −5 10 50 100
p(x) 1c 1c 8c 10c

Where c is a constant.

(a) What is the value of c?

(b) What is P (X > 21)?
(c) What is P (X > 21|X < 60)?

37. Consider the following discrete probability distribution for a random variable X.
Note that the largest value of X and its corresponding probability are missing.

x 10 20 30 ?
p(x) 0.1 0.2 0.3 ?

If E(X) = 60, what is the missing value of X?

38. The probability distribution for a random variable is given by

x 10 11 12 13
P (X = x) 0.4 0.3 0.2 ?

(a) What is the value of the missing probability?

(b) What is P (X > 11)?
(c) What is the mean of this distribution?
(d) What is the variance?

39. Consider again the discrete probability distribution given in Question 38. Suppose
5 values are randomly sampled from this distribution.

(a) What is the probability that exactly three 10s are selected?
(b) What is the probability that at least one 10 is selected?
(c) Values are sampled from this distribution repeatedly. What is the probability
that the frst 11 appears after the sixth sampled value?

40. Consider again the discrete probability distribution given in Question 38. Suppose
20 values are randomly sampled from this distribution.

(a) What is the probability that the number of 10s is less than 8?
Balka ISE 1.09 5.11. CHAPTER EXERCISES 68

(b) What is the probability that the number of 10s is between 2 and 6 (inclusive)?
(c) What is the expected number of 10s?

41. Let X represent the number of heads that come up when a fair coin is tossed 20
times. Let Y represent the number of fours that come up when a six-sided die is
tossed 18 times. It is reasonable to assume that X and Y are independent.

(a) What is the expected value of X + Y ?

(b) What is the standard deviation of X − Y ?

42. Suppose a certain type of lottery ticket costs $1. Let the random variable X repre-
sent the payout on a single ticket. The distribution of X is given in the following
table.
x 0 1 10 100
p(x) 0.689 0.30 0.010 0.001

(a) Find the probability of making a net proft (including the cost of the ticket).
(b) If two tickets are purchased, and the tickets can be assumed to be independent,
what is the probability the two tickets have a combined payout of $100?
(c) What is the expected value of the payout?
(d) What is the standard deviation of the payout?
(e) Given the payout is non-zero, what is the probability the payout is $100?
(f) What is the lottery’s expected proft if 1000 of these tickets are sold?

5.11.2 Concepts

43. Show that E[(X − µ)2 ] = E(X 2 ) − [E(X)]2 . (You may use the property that
E(a + bX) = a + bE(X).)

44. Is the following statement true? Justify your response.

Since E(X 2 ) = E(XX) = E(X)E(X), X is independent of itself.

45. Are the following statements true or false?

(a) If X and Y are two random variables that both have a mean of 0, X and Y
must be independent.
(b) If E(X) > E(Y ), then V ar(X) > V ar(Y ).
(c) If E(X) = 1, then E(X 2 ) = 1

46. Suppose that X has a binomial distribution with n = 15 and p = 0.2, Y has a
Poisson distribution with λ = 6, and Z has an unknown distribution with a mean
of 5 and standard deviation of 18. Suppose that X, Y , and Z are all independent.
Balka ISE 1.09 5.11. CHAPTER EXERCISES 69

(a) What is E(X + Y + Z)?

(b) What is the standard deviation of X + Y − Z?
(c) Could the value of X + Y − Z possibly be negative?

47. Fifteen children are let out during recess onto a playground that has an area of 1000
square metres. Each child detests every other child, and so they space themselves
out as far as possible on the playground. Suppose a 1 square metre area of this
playground is randomly selected. The number of children in the selected area is a
random variable that has which one of the following distributions?

(a) A binomial distribution with n = 15.

(b) A Poisson distribution.
(c) A geometric distribution
(d) A continuous distribution.
(e) None of the above.

48. Tom has a job of completing a delicate fnishing procedure during the production
of expensive pieces of high-tech equipment. If Tom makes a mistake, the equipment
will be destroyed (at a high cost to the manufacturer). Tom is good at his job and on
average has a probability of 0.001 of destroying an individual piece of equipment.
But Tom is a bit shaky psychologically, and if he destroys a piece of equipment
he gets rattled and is more likely to destroy other pieces he works with. Let X
represent the number of pieces Tom destroys in the next 20 pieces he works on.
Then X:

(a) Has a binomial distribution with n = 20 and p = 0.001.

(b) Has a Poisson distribution with λ = 0.02.
(c) Is a discrete random variable, but does not have a binomial or Poisson distri-
bution.
(d) Is not a discrete random variable.
(e) None of the above.

49. Which of the following statements are true?

(a) A discrete random variable can take on a countable number of possible values.
(b) A discrete random variable always represents a count.
(c) The expected value of a discrete random variable must equal one of the vari-
able’s possible values.
(d) The binomial distribution has two parameters: n and p.
(e) A binomial random variable can take on one of n + 1 possible values.

50. Which of the following statements are true?

(a) The variance and standard deviation of a discrete random variable cannot be
Balka ISE 1.09 5.11. CHAPTER EXERCISES 70

equal.
(b) The mean of a random variable cannot be negative.
(c) The standard deviation of a random variable cannot be negative.
(d) If X is a discrete random variable, then Y = 2X + 3 is a discrete random
variable.

51. Which of the following statements are true?

(a) In a binomial setting, the probability of success usually changes from trial to
trial.
(b) The mean of a Poisson random variable can be negative.
(c) If X represents the number of students in a randomly selected 100 m2 area of
a university campus at 3:00 pm on a Tuesday afternoon, then X has a Poisson
distribution.
(d) If p > 0, then the variance of a binomial random variable is less than its mean.
(e) If X is a binomial random variable with n = 10 and p = 0.1, and Y is a
Poisson random variable with λ = 2, then X − Y has a binomial distribution.

5.11.3 Applications

52. The insect Megamelus scutellaris can feed on the sap of the water hyacinth, and
it has been used as a method of biocontrol for this invasive plant species. M.
scutellaris mates on the water hyacinth, and females create oviposition scars on the
plant, laying one or more eggs in each scar. The approximate distribution of the
number of eggs per scar is given in the following table. (Loosely based on a study
by Sosa et al. (2005).)

x 1 2 3 4
p(x) 0.38 0.52 0.05 0.05

(a) What is the expected value of the number of eggs in a randomly selected
oviposition scar? In other words, if we let X represent the number of eggs,
what is E(X)?
(b) What is the standard deviation of the number of eggs?
(c) What is the expected value of the natural log of the number of eggs? That is,
what is E[ln(X)]?
(d) What is E(X 2 )?

53. The Ontario Lottery Corporation runs a “Daily Keno” lottery in which 20 numbers
are randomly selected without replacement from the integers 1 through 70. Before
the drawing, keno players choose a set of numbers (anywhere from 2 to 10 numbers),
and their ticket is a winner if enough of their numbers are randomly chosen. The
Balka ISE 1.09 5.11. CHAPTER EXERCISES 71

following table represents the possible payouts and their probabilities of occurring
for a $1 six-number ticket.2

Number correct 4 or fewer 5 6

Payout 0 $25 $1000
Probability 0.9937921 0.0059123 0.0002956

(a) Verify the probabilities given in the table.

(b) What is the expectation of the payout on a $1 six-number ticket?
(c) What is the standard deviation of the payout on a $1 six-number ticket?
(d) For every $1 spent on this type of ticket, on average how much is the govern-
ment keeping?

54. Consider again the information in Question 53. Suppose that you buy a single
six-number keno ticket in each of the next 50 drawings.

(a) What is the probability that exactly one ticket wins $25?
(b) What is the probability that at least one ticket wins $25?
(c) What is the probability that at least one ticket wins a prize?
(d) What is the expected number of tickets that will win a prize?
(e) What is the standard deviation of the number that will win a prize?

55. You have an agreement with a cheese supplier, which states that he will sell you
1,000 kg of mozzarella cheese for $7000. The supplier is a bit of a gambler and on
March 10 o˙ers you an alternative arrangement: If it does not snow on April 10
you pay only $5,000, but if it snows that day you pay $100,000 instead. Suppose
the probability of snowfall on April 10 is 0.02.

(a) If you choose the alternative arrangement, how much will you owe the supplier
on average?
(b) Which deal is cheaper on average?
(c) Give at least one reason why it is not always wise to choose the option that
results in the greatest expected value.

56. A manufacturer needs to ship an expensive part across the country. The part is
very fragile, and if it breaks it will cost the manufacturer $25,000. A company o˙ers
insurance against breakage at a cost of $1,000.

(a) If the probability of breakage is 0.01, what is the expected proft for the insur-
ance company?
(b) What is the standard deviation of the insurance company’s payout?
2
The listed payouts were extracted from http://www.olg.ca/lotteries/games/howtoplay.do?game=
daily_keno#pt on January 23, 2015. The probabilities have been rounded to the 7th decimal place.
Balka ISE 1.09 5.11. CHAPTER EXERCISES 72

(c) What probability of breakage makes this a break-even proposition for the
manufacturer?

57. People in need of a bone marrow transplant need to fnd a willing donor that
is closely matched in human leukocyte antigens (HLA). The genes that govern
HLA are located on a single chromosome (chromosome 6). Since chromosomes are
inherited from a person’s father and mother, any full sibling of a person has a 14
chance of being an HLA-identical match.

Suppose that a person in need of a bone marrow transplant has 6 siblings (none of
whom are the result of a multiple birth).

(a) What is the distribution of the number of their siblings that are an HLA-
identical match?
(b) What is the probability that at least one of their siblings is an HLA-identical
match?
(c) What is the probability that at least two of their siblings are an HLA-identical
match?
(d) What is the mean number of siblings that are an HLA-identical match?

58. For some types of surveys conducted by random digit dialling, if the call is answered
by an adult at a residence, there is approximately a 9% chance that the person will
complete the survey.3 Suppose you work at a call centre that conducts a survey of
this type.

For the following questions, “call” refers to a call that is answered by an adult at a
residence.

(a) In the next 100 calls, what is the probability that exactly 8 people complete
the survey?
(b) In the next 100 calls, what is the probability that 8, 9, or 10 people complete
the survey?
(c) In the next 100 calls, what is the probability that at least 2 people complete
the survey?
(d) What is the mean number of surveys that will be completed in the next 100
calls ?
(e) What is the standard deviation of the number of surveys that will be completed
in the next 100 calls ?

59. Consider again the information in Question 58. Suppose you will get to leave for
the day once you get the survey completed by 4 adults.
3
Estimated from information on the Public Works and Government Services Canada website http://
www.tpsgc-pwgsc.gc.ca/rop-por/rapports-reports/telephone/introduction-eng.html. Accessed
January 2015.
Balka ISE 1.09 5.11. CHAPTER EXERCISES 73

For the following questions, “call” refers to a call that is answered by an adult at a
residence.

(a) What is the probability that the 4th completed survey occurs on the 50th call?
(b) What is the probability that the 4th completed survey occurs on 50th or 51st
call?
(c) What is the mean number of calls needed to get 4 completed surveys?
(d) What is the standard deviation of the number of calls needed to get 4 com-
pleted surveys?
(e) Challenge: What is the probability it will take more than 100 calls to get the
4 completed surveys?

60. A 2010 investigation by Consumer Reports studied fresh whole broiler chicken at
U.S. retailers. Among other things, they found that salmonella was detectable in
14% of chickens. For the purposes of this question, assume that this fgure is cor-
rect, and 14% of chickens available at retailers have detectable levels of salmonella
contamination.

(a) If 12 chickens are randomly selected from U.S. retailers, what is the probability
that exactly 2 have detectable levels of salmonella?
(b) If 12 chickens are randomly selected from U.S. retailers, what is the probability
that no more than 2 have detectable levels of salmonella?
(c) Suppose a retailer claims that on average, no more than 2% of their chickens
have salmonella. In a random sample of 20 chickens from this retailer, 3
chickens are found to have salmonella contamination. If the retailer’s claim
is true, what is the probability of seeing at least 3 chickens with salmonella
contamination in a sample of 20 chickens?

61. Among eligible blood donors in Canada, approximately 10% have given blood at
some point. Suppose we draw a random sample of 5 eligible blood donors in Canada.

(a) What is the distribution of the number that have given blood at some point?
(b) What is the probability exactly one has given blood at some point?
(c) What is the probability that at least 2 have given blood at some point?
(d) What is the probability that no more than 2 have given blood at some point?
(e) If at least one of these fve people has given blood at some point, what is the
probability that exactly one of these 5 has given blood at some point?

62. You have been giving a friend of yours $1 a week for the past three years (156 weeks)
to buy lottery tickets. Each week, your friend has supposedly been contributing
$1, buying a $2 Lotto 6/49 ticket, checking the numbers, and sharing the wins
and losses with you. You have not won a cash prize in 156 weeks, and you are
growing suspicious of your friend. (According to the Ontario Lottery and Gaming
Corporation, the probability of winning a cash prize in a single Lotto 6/49 drawing
Balka ISE 1.09 5.11. CHAPTER EXERCISES 74

is approximately 0.03.)

If one ticket is purchased per week for 156 weeks:

(a) What is the distribution of the number of tickets that win a cash prize?
(b) What is the expected number of tickets that win a cash prize?
(c) What is the probability that no ticket wins a cash prize?
(d) Is there strong evidence that your friend has been ripping you o˙?

63. Suppose that the number of fatal crashes per year involving U.S. commercial airlines
has a mean of approximately 1.1.4 To a reasonable approximation, fatal crashes
can be thought of as occurring randomly and independently.

(a) What is the probability that there are no fatal crashes involving U.S. commer-
cial airlines in the next year?
(b) What is the probability that there are 2 or 3 fatal crashes involving U.S.
commercial airlines in the next year?
(c) What is the probability that there are more than 4 fatal crashes involving U.S.
commercial airlines in the next year?
(d) What is the standard deviation of the number of fatal crashes involving U.S.
commercial airlines in the next year?

64. Consider again the information in Question 63.

(a) What is the probability that there is at least 1 fatal crash involving U.S.
commercial airlines in the next 2 year period?
(b) What is the probability that there is more than 1 fatal crash involving U.S.
commercial airlines in the next 2 year period?
(c) What is the mean number of fatal crashes involving U.S. commercial airlines
in the next 2 year period?
(d) What is the standard deviation of the number of fatal crashes involving U.S.
commercial airlines in the next 2 year period?

65. According to Wikipedia, objects with a diameter in excess of 1 km hit the earth
at a rate of approximately two every million years. Assume that the number of
impacts in any given time period has (approximately) a Poisson distribution.

(a) What is the approximate probability that there is at least one impact of this
size in a given 100,000 year period?
4
This value is the mean number of annual fatal crashes for the period 1990-
2012, calculated from data obtained from the National Transportation Safety Board
(http://www.ntsb.gov/investigations/data/Pages/paxfatal.aspx). It is likely not very far from
the true value. Let’s assume it to be correct for the purposes of this question.
Balka ISE 1.09 5.11. CHAPTER EXERCISES 75

(b) What is the approximate probability that there is at least one impact of this
size in a given 10,000 year period?

66. In a famous series of genetics experiments, Gregor Mendel investigated inheritance

in sweet peas. In one experiment, he crossed one pure line of peas that had round
yellow seeds with another pure line that had wrinkled green peas. A round shape
and yellow colour are dominant traits, so the peas resulting from the frst generation
cross all had round yellow seeds (they had the round/yellow phenotype). This
frst generation was then self-crossed. For the second generation, under Mendelian
inheritance with independent assortment, a 9:3:3:1 ratio of phenotypes would be
expected. In other words, the distribution of phenotypes would be:

Phenotype Round/Yellow Round/Green Wrinkled/Yellow Wrinkled/Green

Probability 9
16
3
16
3
16
1
16

Suppose we examine 50 pea plants from this type of cross.

(a) What is the distribution of the number of round/yellow plants?

(b) What is the mean number of round/yellow plants?
(c) What is the standard deviation of the number of round/yellow plants?
(d) What is the probability that there are exactly 4 wrinkled/green plants?
(e) What is the probability that there are 28 round/yellow plants, 12 round/green
plants, 6 wrinkled/yellow plants, and 4 wrinkled/green plants?

5.11.4 Extra Practice Questions

67. Suppose that for a binomial random variable X, n = 10 and µ = 2.4.

(a) What is p?
(b) What is P (X = 2)?
(c) What is σ?

68. A certain coin sold by a novelty store has a probability of coming up heads of 23 .
Let X be a random variable representing the total number of heads when this coin
is tossed twice.

(a) What is the mean of X?

(b) What is the standard deviation of X?

69. According to the Pocket World in Figures by the Economist, 17.8% of the Canadian
population is over 60 years old. Suppose 20 Canadians are randomly selected.
Balka ISE 1.09 5.11. CHAPTER EXERCISES 76

(a) What is the distribution of the number of people in the sample that are over
60 years old?
(b) What is the probability that at least 2 are over 60 years old?
(c) What is the probability that exactly 2 are over 60 years old?
(d) What is the probability that none are over 60 years old?
(e) What is the expected number of people that are over 60 years old?

70. Approximately 60% of Canadian adults voted in the last federal election. Suppose
we draw a random sample of fve people who were Canadian adults at the time of
the last election.

(a) What is the distribution of the number of people in the sample that voted in
the last federal election?
(b) What is the probability that exactly 4 voted in the last federal election?
(c) What is the probability that four or more voted in the last federal election?
(d) Given at least four voted, what is the conditional probability that all 5 voted?
(e) What is the mean of the number that voted in the last election?

71. A 2011 report by Statistics Canada showed the unemployment rate in Canada was
approximately 7.6% (7.6% of the labour force was out of work). If 14 members of
the Canadian labour force were randomly selected during this time period:

(a) What is the probability exactly three were out of work?

(b) What is the probability more than 2 were out of work?
(c) Given at least 2 of the 14 were out of work, what is the probability at least 1
of the 14 was out of work?
(d) Given at least 3 of the 14 were out of work, what is the probability no more
than one of the 14 was out of work?

72. Approximately 35% of cars that are at least 13 years old fail the Ontario Drive
Clean emissions test. Suppose we draw a random sample of 10 emissions tests on
cars that are at least 13 years old.

(a) What is the probability that exactly 2 fail?

(b) What is the probability that no more than 2 fail?
(c) What is the probability that more than 2 fail?
(d) What is the expectation of the number of failures?
(e) What is the standard deviation of the number of failures?
(f) Suppose a certain garage has given a failing emissions grade to each of the 10
cars they have tested that are at least 13 years old. Is this strong evidence the
garage is fraudulent?

73. A certain amount of a material is experiencing radioactive decay, with a decay

Balka ISE 1.09 5.11. CHAPTER EXERCISES 77

rate of 0.8 per second (an average of 0.8 radioactive decays per second). It is
reasonable to assume the number of decays in a given time period will follow a
Poisson distribution.

(a) In a one second period, what is the probability the material has exactly two
radioactive decays?
(b) In a three second period, what is the probability the material has exactly two
radioactive decays?
(c) In a three second period, what is the probability the material has no more
than two radioactive decays?

74. At a wood chipper in a paper mill, log jams occur (approximately) randomly and
independently, at the rate of 0.2 per hour.

(a) What is the probability that there are exactly 3 log jams in an 8 hour day?
(b) What is the probability that there is at least one log jam in an 8 hour day?
(c) Challenge: In 12 randomly selected hours, what is the probability that there
is at least one log jam in at least 4 of the hours?.

75. Airplanes arrive at a small airport at an average rate of 5 per hour. Suppose
the number of planes arriving in any given period approximately follows a Poisson
distribution.

(a) What is the probability that in a 2 hour period, at least one plane arrives?
(b) What is the probability that in a 3 hour period, exactly 3 arrive?
(c) What is the probability that in a half-hour period, no planes arrive?
(d) What is the standard deviation of the number of planes that arrive in an 8
hour period?

76. Suppose that planes at a remote airport are very poorly maintained, and have a
0.01 probability of crash landing before their destination. Assuming independence,
in the next 10 planes that take o˙:

(a) What is the distribution of the number that crash?

(b) What is the probability that at least one crashes?
(c) What is the probability that exactly one crashes?
(d) What is the probability they all crash?
(e) What is the standard deviation of the number that crash?

77. The probability that a randomly selected Canadian is a medical doctor is approxi-
mately 0.002.

(a) If Canadians are randomly selected, what is the probability the fourth person
selected is the frst that is a medical doctor?
Balka ISE 1.09 5.11. CHAPTER EXERCISES 78

(b) What is the probability that more than 1000 Canadians must sampled before
encountering the frst medical doctor?
(c) If 80 Canadians are randomly selected, what is the probability that exactly
one is a medical doctor?
(d) If 80 Canadians are randomly selected, what is the probability that more than
two are medical doctors?

78. A company is interested in purchasing a new automatic welding machine. The

producer of the machine claims that for a certain type of weld, each weld has a 90%
chance of being successful.

Suppose that the company interested in the welder decides to test it by having it
perform welds until it fails for the frst time. Assume that the producer’s claim
is true (each weld will be successful with probability 0.90 and unsuccessful with
probability 0.10), and that the trials are independent.

(a) What is the distribution of the number of welds needed to get the frst unsuc-
cessful weld?
(b) What is the probability that the frst unsuccessful weld occurs on the third
trial?
(c) What is the probability that the frst unsuccessful weld occurs on or before
the third trial?
(d) What is the probability that the frst unsuccessful weld occurs after the 13th
trial?

79. Consider again the information in Question 78. Suppose the company purchases the
machine and uses it whenever a weld of the appropriate type is required. (Assume
that the probability of a successful weld is always 0.90 and that the trials are
independent.)

(a) What is the probability that the third unsuccessful weld occurs on the 22nd
welding job?
(b) In the next 20 welds, what is the probability that exactly 3 are unsuccessful?
(c) In the next 20 welds, what is the probability that at least 3 are unsuccessful?
(d) What is the mean number of unsuccessful welds in a sample of 20 welds?

80. Suppose that you get a stimulating job inspecting shipments of light bulbs that
arrive at a factory. Your job is to take out one randomly selected bulb and test
it. If the bulb works, the shipment is accepted. If the bulb malfunctions then
the shipment is rejected. Unknown to you, in reality the probability that any
randomly selected bulb malfunctions is 0.04, and the shipments can be considered
to be independent.

(a) What is the probability that the frst shipment you must reject occurs on the
Balka ISE 1.09 5.11. CHAPTER EXERCISES 79

10th shipment?
(b) What is the probability that the frst shipment you must reject occurs after
the 10th shipment?
(c) In the frst 60 shipments, what is the probability you must send back no more
than one shipment?

81. A room contains 30 women and 55 men, and 5 of these people are randomly selected
without replacement,

(a) What is the exact probability of getting exactly 4 men in the sample?
(b) Use the binomial distribution to approximate the probability of getting 4 men
in the sample.

82. Suppose the distribution of dandelions in a large meadow is approximately Poisson,

with a mean of 4 dandelions per square metre. A 1 m2 area in this meadow is
randomly selected.

(a) What is the probability there are exactly 2 dandelions in the area?
(b) What is the probability there are more than 2 dandelions in the area?
(c) Given there is at least one dandelion in the area, what is the probability there
are exactly three dandelions?
(d) What is the standard deviation of the number of dandelions in the randomly
selected area?

83. Suppose you design a new type of dental implant. You feel that if the implant
is placed in a randomly selected dental patient, the probability that it does not
take properly is 0.05. As part of a test of the new implant, you intend to place
your implant into patients until you get 3 unsuccessful implants. (Assume for
this question that your probability assessment is correct, and also assume that the
patients can be considered independent.)

(a) What is the probability that the frst unsuccessful implant occurs on the fourth
trial?
(b) What is the probability that the frst unsuccessful implant occurs after the
11th trial?
(c) What is the probability that the third unsuccessful implant occurs on the 28th
trial?

84. A dollar store bin contains 15 rubber footballs, 20 rubber soccer balls, and 8 rubber
volleyballs. Suppose 6 of these balls are randomly selected without replacement.

(a) What is the probability that exactly 2 footballs are selected?

(b) What is the expectation of the number of footballs selected?
(c) What is the probability that no more than 1 soccer ball is selected?
Balka ISE 1.09 5.11. CHAPTER EXERCISES 80

(d) What is the probability 3 footballs, 2 soccer balls, and 1 volleyball are selected?

85. A room contains 5 biomedical engineers, 7 computer engineers, 5 mechanical engi-

neers, and 3 water resource engineers.

(a) If 6 people are randomly selected without replacement, what is the probability
there are exactly 3 computer engineers selected in the group of 6?
(b) If 6 people are randomly selected without replacement, what is the probability
there are exactly 2 biomedical engineers, 2 computer engineers, 1 mechanical
engineer, and 1 water resource engineer in the sample?

86. A rental car company has 20 cars available: 4 Hondas, 6 Fords, 3 Toyotas, and
7 Chrysler vehicles. If they randomly pick 5 of these cars for inspection (without
replacement), what is the probability the sample consists of 1 Honda, 1 Ford, 1
Toyota, and 2 Chrysler vehicles?

87. Consider the following distribution of a categorical variable.

Category A B C D
Probability 0.4 0.3 0.2 0.1

Suppose 5 values are sampled independently from this distribution. What is the
probability there will be three A’s, one B, and one C?
Chapter 6

Continuous Random Variables and

Continuous Probability
Distributions

J.B.’s strongly suggested exercises: 1, 5, 6, 8, 9, 10, 11, 13, 14, 15, 21, 23, 24, 27

6.1 Introduction

6.2 Properties of Continuous Probability Distributions

1. Test your conceptual understanding: Which of the following statements are true?
(You should be able to explain why a statement is true or false.)

(a) For continuous random variables, probabilities correspond to areas under the
density curve.
(b) The area under any continuous probability distribution is 1.
(c) A probability density function can take on negative values.
(d) A probability density function can take on values greater than 1.
(e) f (x) = P (X = x) for all x.
(f) f (x) = µ whenever the distribution is symmetric.

81
Balka ISE 1.09 6.2. PROPERTIES OF CONTINUOUS PROBABILITY DISTRIBUTIONS 82

6.2.1 An Example Involving Integration

2. Warning: This question requires calculus (a basic integration).

Suppose the random variable X has the probability density function:

cx2 for 0 < x < 3


f (x) =
0 elsewhere

(a) What value of c makes this a legitimate probability distribution?

(b) What is P (X > 2)?
(c) What is P (X < 1)?
(d) What is the 32nd percentile of this distribution?
(e) What is the median of this distribution?
(f) What is the mean of this distribution?

3. Warning: This question requires calculus (a basic integration).

Let the random variable X have the probability density function:
⎧ c
⎨ for 4 < x < 10
f (x) = x2
⎩ 0 elsewhere

(a) What value of c makes this a legitimate pdf?

(b) What is P (8 < X < 9)?
(c) What is the 44th percentile of this distribution?

4. Warning: This question requires calculus—a basic integration.

Suppose the response time of a certain computer system approximately follows an

exponential distribution with a mean of 0.10 seconds:

f (x) = 10e−10x

for x > 0.

(a) What is the probability a randomly selected response time exceeds 0.22 sec-
onds?
(b) What is the probability a randomly selected response time exceeds 0.02 sec-
onds?
(c) What is the probability a randomly selected response time lies between 0.08
and 0.12 seconds?
(d) What is the 25th percentile of response time for this computer system?
(e) What is the median response time for this computer system?
Balka ISE 1.09 6.3. THE CONTINUOUS UNIFORM DISTRIBUTION 83

6.3 The Continuous Uniform Distribution

5. Let X have a continuous uniform distribution on the interval [5, 25].

(a) What is f (x)?

(b) What is P (X = 8.0)?
(c) What is P (X > 8.0)?
(d) What is P (X > 8.0|X < 7.3)?
(e) What is P (X > 8.0|X > 7.3)?
(f) What is the median of this distribution?
(g) What is the mean of this distribution?
(h) What is the 47th percentile of this distribution?
(i) What is the interquartile range of this distribution?

6. Which of the following statements about the continuous uniform distribution are
true? (You should be able to explain why a statement is true or false.)

(a) The uniform distribution is symmetric.

(b) For any uniform distribution, the mean and median are equal.
(c) For any uniform distribution, Q1 = −Q3 .
(d) For any uniform distribution, Q3 − Q2 = Q2 − Q1 .
(e) If a random variable has a uniform distribution, then it cannot take on negative
values.
(f) If a random variable has a uniform distribution, then its standard deviation is
greater than its mean.

6.4 The Normal Distribution

7. What are the parameters of the normal distribution? Can any of these parameters
be negative?

8. Which of the following statements are true?

(a) The normal distribution is symmetric about µ.

(b) If X is a normally distributed random variable, then P (X < 0) > 0.
(c) If X is a normally distributed random variable, then P (X = 0) = 0.
(d) The standard deviation of a normal distribution cannot be greater than the
mean.
(e) The mean and median of a normal distribution are always equal.
Balka ISE 1.09 6.4. THE NORMAL DISTRIBUTION 84

6.4.1 Finding Areas Under the Standard Normal Curve

9. Suppose the random variable Z has the standard normal distribution.

(a) What is P (Z < 0.37)?

(b) What is P (Z > 0.37)?
(c) What is P (0 < Z < 0.37)?
(d) What is P (Z < −1.28)?
(e) What is P (Z > −1.28)?
(f) What is P (Z = −1.28)?
(g) What is P (−1.28 < Z < 0)?
(h) What is P (Z > 2.41|Z > 0)?
(i) What is P (−2.41 < Z < 2.41|Z > 0)?
(j) What is the 83rd percentile of the standard normal distribution?
(k) What is the interquartile range of the standard normal distribution?
(l) If two values are independently sampled from the standard normal distribution,
what is the probability they are both less than 2.13?

10. Suppose the random variable Z has the standard normal distribution.

(a) What is P (Z > −2.73)?

(b) What is P (Z > 2.73)?
(c) What is the value z such that P (Z > z) = 0.1210?
(d) What is the value z such that P (Z < z) = 0.1210?
(e) What is the value z such that P (−z < Z < z) = 0.95?
(f) What is the value of z such that P (Z > z) = 0.7611?
(g) If two values are sampled independently from the standard normal distribution,
what is the probability that both values are greater than 2.00?

11. P (Z > 31788905573847358927509487) is closest to which one of the following?

(a) 0
(b) 0.5
(c) 1
(d) 31788905573847358927509487
(e) ∞

6.4.2 Standardizing Normally Distributed Random Variables

12. Suppose the random variable X has a normal distribution with a mean of µ = 120
and a standard deviation of σ = 20.
Balka ISE 1.09 6.4. THE NORMAL DISTRIBUTION 85

(a) What is P (X < 105)?

(b) What is P (X > 87)?
(c) What is P (92 < X < 108)?
(d) What is P (82 < X < 87)?
(e) What is P (X = 84)?
(f) What is the 20th percentile of the distribution of X?
(g) What is the 80th percentile of the distribution of X?
(h) What is the interquartile range of the distribution of X?

13. A study1 found that birth weights of female African elephants born in captivity are
approximately normally distributed with a mean of 95.1 kg and a standard deviation
of 13.7 kg. (The following questions are all in reference to African elephants born in
captivity. The distribution of weights of elephants born in the wild is likely di˙erent
from that of those born in captivity.)

(a) What is the probability that a randomly selected female newborn African
elephant weighs more than 120 kg?
(b) What is the probability that a randomly selected female newborn African
elephant weighs less than 100 kg?
(c) What is the probability that a randomly selected female newborn African
elephant weighs between 90.0 and 110.0 kg?
(d) What is the 30th percentile of birth weights of female African elephants?
(e) What is the 80th percentile of birth weights of female African elephants?

14. Test your conceptual understanding: Which of the following statements are true?
(You should be able to explain why a statement is true or false.)

(a) If X has a normal distribution with mean µ and standard deviation σ, then
X −µ
σ has the standard normal distribution.
(b) If Z has the standard normal distribution, then the mean and standard devi-
ation of Z are equal.
(c) The frst quartile of the standard normal distribution is greater than 0.
(d) The frst quartile of any normal distribution is greater than 0.
(e) If X has a normal distribution, then P (X > µ) = P (X < µ).
1
Dale, R. (2010). Birth statistics for African (Loxodonta africana) and Asian (Elephas maximus) ele-
phants in human care: History and implications for elephant welfare. Zoo Biology, 29:87–103. (The
mean and standard deviation given in this question are estimates based on this article, but assume that
they are the true values for the purposes of this question.)
Balka ISE 1.09 6.5. NORMAL QUANTILE-QUANTILE PLOTS 86

6.5 Normal Quantile-Quantile Plots: Is the Data Ap-

proximately Normality Distributed?

6.5.1 Examples of Normal QQ Plots for Di˙erent

Distributions

15. Consider the plots Figure 6.1, which represent normal quantile-quantile plots for 4
di˙erent samples.
Sample Quantiles

Sample Quantiles

Theoretical Quantiles Theoretical Quantiles

(a) A (b) B
Sample Quantiles

Sample Quantiles

Theoretical Quantiles Theoretical Quantiles

(c) C (d) D

Figure 6.1: Four normal QQ plots.

(a) What does Plot A tell us about the shape of the distribution from which this
sample is drawn?
(b) What does Plot B tell us about the shape of the distribution from which this
sample is drawn?
(c) What does Plot C tell us about the shape of the distribution from which this
sample is drawn?
(d) What does Plot D tell us about the shape of the distribution from which this
sample is drawn?

16. Draw a sketch of a normal quantile-quantile plot in which the data shows strong
left skewness, but has one large outlier (one value that is much larger than would
be expected under normality).
Balka ISE 1.09 6.6. OTHER IMPORTANT CONTINUOUS PROBABILITY DISTRIBUTIONS 87

6.6 Other Important Continuous Probability Distribu-

tions

6.6.1 The χ2 Distribution

6.6.2 The t Distribution

6.6.3 The F Distribution

6.7 Chapter Exercises

6.7.1 Basic Calculations

17. Suppose random variable X has the uniform distribution on the interval [0, 50].

(a) What is P (X < 10)?

(b) What is P (X > 10)?
(c) What is P (12 < X < 23)?
(d) What is P (−5 < X < 5)?
(e) What is the value a such that P (X > a) = 0.80?
(f) What is the 20th percentile of this distribution?
(g) What is the median of this distribution?

18. Consider the continuous probability distribution illustrated in Figure 6.2.

f(x) ?

0
0.0 0.5 1.0
x

Figure 6.2: A continuous probability distribution.

(a) What is the height of the curve at x = 1?

(b) What is the height of the curve at x = 0.50?
(c) What is the probability density function?
Balka ISE 1.09 6.7. CHAPTER EXERCISES 88

(d) What is the height of the curve at x = 2?

(e) What is P (X < 1.5)?
(f) What is P (X < 0.5)?
(g) What is P (X > 0.5)?
(h) What is the median of this distribution?

6.7.2 Concepts

19. P (Z > −458958643961346) is closest to which one of the following?

(a) 0
(b) 0.5
(c) 1
(d) −458958643961346
(e) ∞

20. Test your conceptual understanding: Which of the following statements are true?
(You should be able to explain why a statement is true or false.)

(a) The 95th percentile of a continuous random variable cannot be less than 0.
(b) The mean of a continuous random variable cannot be negative.
(c) The mean of a continuous random variable can be greater than the third
quartile.
(d) The normal distribution is symmetric.
(e) If X has a normal distribution, then the mean and median of X are equal.

6.7.3 Applications

21. A study found that the total serum cholesterol level in Canadian men aged 60–79
is approximately normally distributed with a mean of 4.7 mmol/L and a standard
deviation of 1.2 mmol/L.2

(a) High total serum cholesterol levels have been linked to heart disease. By one
standard, a total serum cholesterol level of at least 6.2 mmol/L is considered
high. If one Canadian man in this age group is randomly selected, what is the
probability his total serum cholesterol level is at least 6.2 mmol/L?
2
The mean and standard deviation are estimates based on information from the Canadian Health Mea-
sures Survey (Statistics Canada, 2013). Assume these are the true values for the purposes of this
question.
Balka ISE 1.09 6.7. CHAPTER EXERCISES 89

(b) What proportion of Canadian men in this age group have high total serum
cholesterol? (As judged by the 6.2 mmol/L guideline.)
(c) The 6.2 mmol/L cuto˙ value for high total serum cholesterol is quite high,
and some sources suggest that a person should aim to keep their total serum
cholesterol below 4.6 mmol/L. What proportion of Canadian males between
60 and 79 years of age have a total serum cholesterol level that is below 4.6
mmol/L?
(d) What proportion of Canadian males between 60 and 79 years of age have total
serum cholesterol levels between 4.6 mmol/L and 6.2 mmol/L?
(e) What is the 20th percentile of total serum cholesterol levels for Canadian men
in this age group?
(f) What is the 80th percentile of total serum cholesterol levels for Canadian men
in this age group?

22. A study found that heights of Canadian females aged 20 to 39 are approximately
normally distributed with a mean of 163.3 cm and a standard deviation of 6.4 cm.
(See footnote 2.)

(a) What is the probability a randomly selected Canadian female in this age group
is at least 6 feet (182.88 cm) tall?
(b) What is the probability a randomly selected Canadian female in this age group
is taller than 170.0 cm?
(c) What is the probability a randomly selected Canadian female in this age group
is shorter than 170.0 cm?
(d) What is the probability a randomly selected Canadian female in this age group
is taller than 150.0 cm?
(e) What is the probability a randomly selected Canadian female in this age group
has a height that is within 1 standard deviation of the mean height?
(f) What is the 20th percentile of heights of adult Canadian females in this age
group?

23. A study found that heights of Canadian males aged 20 to 39 are approximately
normally distributed with a mean of 177.7 cm and a standard deviation of 6.5 cm.
(See footnote 2.)

(a) What is the probability a randomly selected Canadian male in this age group
is at least 6 feet (182.88 cm) tall?
(b) What is the probability a randomly selected Canadian male in this age group
is taller than 170.0 cm?
(c) What is the probability a randomly selected Canadian male in this age group
is shorter than 170.0 cm?
(d) What is the 99th percentile of heights of Canadian males in this age group?
(e) If 3 Canadian males in this age group are randomly selected, what is the
Balka ISE 1.09 6.7. CHAPTER EXERCISES 90

probability they are all taller than 170.0 cm?

(f) If 2 Canadian males in this age group are randomly selected, what is the
probability they are both at least 6 feet (182.88 cm) tall?
(g) If 5 Canadian males in this age group are randomly selected, what is the
probability that exactly 2 of them are at least 6 feet (182.88 cm) tall?

24. Sea urchin gonads are a popular seafood in Japan. A type of green sea urchin
(Hemicentrotus pulcherrimus) is harvested o˙ the coast of Japan, but those har-
vested o˙ the northeast coast are often found to have a bitter taste and are not
commercially viable. A study3 investigated characteristics of green sea urchins har-
vested o˙ the northeast coast. In one part of the study, it was found that the test
diameters of mature male green sea urchins in this area are approximately normally
distributed with a mean of 37.9 mm and a standard deviation of 3.8 mm. (The test
is the sea urchin’s shell.)

(a) What is the probability that a randomly selected mature male green sea urchin
in this area has a test diameter that is greater than 40.0 mm?
(b) What proportion of mature male green sea urchins in this area have a test
diameter that is less than 30.0 mm?
(c) What proportion of mature male green sea urchins in this area have test di-
ameters that are between 30.0 mm and 40 mm?
(d) What is the median test diameter of mature male green sea urchins in this
area?
(e) What is the 25th percentile of test diameters of mature male green sea urchins
in this area?

25. A study4 investigated the weight of cereal in bags of cereal purchased at a grocery
store. A certain brand of cereal had a size of bag that had a nominal (stated)
weight of 368 grams. It was found that the actual weight of cereal in these bags
was approximately normally distributed with a mean of 374.5 grams and a standard
deviation of 2.8 grams.

(a) What proportion of bags of this type contain at least 380 grams of cereal?
(b) What proportion of bags of this type contain less than the stated amount of
cereal?
(c) A bag of this type is randomly selected. What is the probability that it contains
between 370 and 380 grams of cereal?
(d) What is the 50th percentile of the weight of cereal in bags of this type?
3
Murata, Y. (2001). Studies on a novel bitter amino acid, pulcherrimine in the green sea urchin gonads.
PhD thesis, Kyoto University. (The mean and standard deviation in this question are estimates based
on this article, but assume that they are the true values for the purposes of this question.)
4
Personal study conducted by your author. (The mean and standard deviation in this question are
estimates based on this study, but assume that they are the true values for the purposes of this question.)
Balka ISE 1.09 6.7. CHAPTER EXERCISES 91

(e) What is the 15th percentile of the weight of cereal in bags of this type?

26. A study5 investigated the amount of protein in a type of 2% cow’s milk. The amount
of protein in a cup (245 grams) of milk was found to be approximately normally
distributed with a mean of approximately 8.0 grams and a standard deviation of
approximately 0.30 grams.

(a) What is the probability that a randomly selected cup of this type of milk
contains less than 8.10 grams of protein?
(b) What is the probability that a randomly selected cup of this type of milk
contains more than 14 grams of protein?
(c) What is the probability that a randomly selected cup of 2% milk contains
between 8.05 and 8.20 grams of protein?
(d) What is the 99th percentile of the amount of protein in a cup of 2% milk?
(e) What is the 75th percentile of the amount of protein in a cup of 2% milk?

27. The total amount of fat in one serving (85 grams) of pan-browned crumbled 90%
lean ground beef is approximately normally distributed with a mean of 10.2 grams
and a standard deviation of 1.8 grams. (See footnote 5.)

(a) What is the probability that a randomly selected serving of this type of beef
contains less than 15.0 grams of fat?
(b) What is the probability that a randomly selected serving of this type of beef
contains between 8.0 and 12.0 grams of fat?
(c) If two servings of this type of beef are randomly selected, what is the proba-
bility that each serving contains more than 14.0 grams of fat?
(d) What is the 15th percentile of the amount of fat in a serving of this type of
beef?
(e) What is the 95th percentile of the amount of fat in a serving of this type of
beef?

28. Suppose that the head lengths of adult female lizards of the species Phrynocephalus
versicolor are approximately normal with a mean of 11.90 mm and a standard
deviation of 0.41 mm.6

(a) If an adult female lizard of this species is randomly selected, what is the
probability its head length is between 11.0 and 12.0 mm?
(b) If an adult female lizard of this species is randomly selected, what is the
probability its head length is at least 11.5 mm?
5
The mean and standard deviation in this question are estimates based on information from the USDA
National Nutrient Database. Assume that they are the true values for the purposes of this question.)
6
The mean and standard deviation are based on a study by Qu et al. (2011). Assume they are the true
values for the purposes of this question.
Balka ISE 1.09 6.7. CHAPTER EXERCISES 92

(c) A person captures a female lizard that they believe to be of this species. The
lizard’s head length is found to be 22.5 mm. Should they rethink their belief
that this lizard is of this species?
(d) What is the 10th percentile of the head lengths for adult female P. versicolor
lizards?
(e) What is the 90th percentile of the head lengths for adult female P. versicolor
lizards?

29. Many studies have investigated the weight of a human organs, often basing the
fndings on post-mortem results from accidental deaths. In one study7 , it was found
that the human heart in adult Caucasian males had a mean of 365.0 grams and a
standard deviation of 71.0 grams. For the purposes of this question, assume that
the heart weight of adult Caucasian males is approximately normally distributed
with a mean of 365.0 grams and a standard deviation of 71.0 grams.

(a) What is the probability a randomly selected adult Caucasian male has a heart
that weighs more than 400.0 grams?
(b) What is the probability a randomly selected adult Caucasian male has a heart
that weighs more than a pound (453.6 grams)?
(c) What is the probability a randomly selected adult Caucasian male has a heart
that weighs less than 300 grams?
(d) What is the probability a randomly selected adult Caucasian male has a heart
that weighs between 300 and 400 grams?
(e) What is the 90th percentile of heart weight for adult Caucasian males?

30. Weights of full-term newborn babies in the United States are (very roughly) nor-
mally distributed with a mean of approximately 3.3 kg and a standard deviation of
approximately 0.5 kg. (This is based on information found in a study by Martin
et al. (2015).)

(a) If one full-term newborn baby is randomly selected, what is the probability it
weighs less than 2.0 kg?
(b) If one full-term newborn baby is randomly selected, what is the probability it
weighs less than 4.0 kg?
(c) What proportion of full-term newborn babies in the United States weigh be-
tween 2.0 and 4.0 kg?
(d) If one full-term newborn baby is randomly selected, what is the probability it
weighs more than 6.0 kg?
(e) What is the 90th percentile of full-term newborn baby weights in the United
States?
7
Lorin de la Grandmaison et al. (2001). Organ weight in 684 adult autopsies: new tables for a Caucasoid
population. Forensic Science International, 119:149–154.
Balka ISE 1.09 6.7. CHAPTER EXERCISES 93

(f) What is the 25th percentile of full-term newborn baby weights in the United
States?

6.7.4 Extra Practice Questions

31. Under certain conditions, the lifetimes of a certain type of insect are approximately
normally distributed with a mean of 60.2 days and a standard deviation of 18.0
days.

(a) If one of these insects is randomly selected, what is the probability it lives
longer than 100 days?
(b) If one of these insects is randomly selected, what is the probability it lives
longer than 200 days?
(c) If one of these insects is randomly selected, what is the probability it dies
before 50 days?
(d) What is the probability a randomly selected insect dies between 50 and 70
days?
(e) What is the 90th percentile of the lifetimes of these insects under these condi-
tions?

32. A supplier of pure oxygen claims that their containers of pure oxygen have a mean
purity of 99.500%, with a standard deviation of 0.048%. In your frst dealing with
this company, your frst container has a purity level of 99.43%. Assume the purity
levels are approximately normally distributed.

(a) Assuming the company’s claim is true, what is the probability that a randomly
selected container has a purity level less than 99.43%?
(b) Assuming the company’s claim is true, what is the probability that a randomly
selected container has a purity level greater than 99.60%?
(c) Assuming the company’s claim is true, what is the 10th percentile of purity
levels?
(d) Assuming the company’s claim is true, what is the interquartile range (IQR)
of purity levels?
(e) Assuming the company’s claim is true, what is the probability that your next
4 containers all have purity levels greater than 99.40%? (Suppose that it is
reasonable to assume the containers are independent.)

33. The weights of a type of adult pigeon in a city are approximately normally dis-
tributed with a mean of 370 grams and a standard deviation of 28 grams.

(a) If one adult pigeon is randomly selected, what is the probability it weighs less
than 350 grams?
Balka ISE 1.09 6.7. CHAPTER EXERCISES 94

(b) If one adult pigeon is randomly selected, what is the probability it weighs more
than 1 pound (453.6 grams)?
(c) If one adult pigeon is randomly selected, what is the probability it weighs
between 350 and 400 grams?
(d) If one adult pigeon is randomly selected, what is the probability it weighs more
than 6.0 kilograms?
(e) What is the 5th percentile of adult pigeon weights in this city?

34. At a certain large pizza restaurant, the daily demand for cheese is approximately
normally distributed with a mean of 50 kg and a variance of 36 kg2 .

(a) On a randomly selected day, what is the probability the restaurant needs more
than 60 kg of cheese?
(b) On a randomly selected day, what is the probability the restaurant needs less
than 45 kg of cheese?
(c) What is the 80th percentile of cheese demand?
(d) If the restaurant wants to be 99% sure that they have enough cheese on any
given day, how much cheese should they have on hand each day?

35. A type of steel shaft is produced by a process that results in a mean diameter of 4.10
cm with a standard deviation of 0.05 cm. The shaft diameters are approximately
normally distributed.

(a) What is the probability that a randomly selected shaft has a diameter in excess
of 4.23 cm?
(b) What is the probability that a randomly selected shaft has a diameter less
than 4.00 cm?
(c) What is the 10th percentile of shaft diameters?
(d) What is the 90th percentile of shaft diameters?
(e) Suppose a shaft is usable if it has a diameter between 4.00 and 4.20 cm,
otherwise it must be scrapped. What proportion of shafts created by this
process must be scrapped?
Chapter 7

Sampling Distributions

J.B.’s strongly suggested exercises: 1, 6, 7, 8, 9, 10, 20, 21, 22, 23, 24, 25

7.1 Introduction

1. Which of the following statements are true?

(a) We view statistics as random variables that have probability distributions.

(b) We view parameters as random variables that have probability distributions.
(c) The sampling distribution of a statistic is the probability distribution of the
statistic.
(d) In repeated sampling, the value of a statistic will vary about the parameter it
estimates.
(e) In repeated sampling, the value of a parameter will vary about the statistic
that estimates it.

2. Suppose we are about to sample 2 values (without replacement) from a population

that contains only 4 values. Unknown to us, the entire population is:

12, 13, 15, 16

(a) What is the mean of the population?

(b) How many possible samples of size n = 2 are there?
(c) List all the possible samples of size 2, their probabilities of occurring, and the
value of the sample mean for each sample.
(d) What is the sampling distribution of the sample mean in this situation?
(e) What is the mean of the sampling distribution of the sample mean?

95
Balka ISE 1.09 7.2. THE SAMPLING DISTRIBUTION OF THE SAMPLE MEAN 96

3. Consider again the information in Question 2.

(a) What is the value of the sample variance for each possible sample?
(b) What is the sampling distribution of the sample variance in this situation?
(c) If we draw a sample of size n = 2 from the population given in Question 2,
what is the probability that the sample variance will equal 0.5?

4. Let X be the the number of heads that come up when a fair coin is tossed once,
and let X
¯ be the mean number of heads per toss when a fair coin is tossed twice.

(a) What is the probability distribution of X?

(b) What is the mean of the probability distribution of X?
(c) What is the variance of the probability distribution of X?
(d) What is the probability distribution of X?
¯
(e) What is the mean of the probability distribution of X?
¯
(f) What is the variance of the probability distribution of X?
¯

5. Tim Horton’s co˙ee shop chain has long held a “Roll up the Rim” contest, where
purchasers of a hot beverage can roll up the rim of the cup, possibly revealing that
the cup is a prize winner. The probability of winning a prize on each cup changes
slightly from year to year, but it is often approximately 16 .

(a) If you purchase 2 cups, what is the distribution of the number of winning
cups?1
(b) If you purchase 2 cups, what is the sampling distribution of the sample pro-
portion of winning cups? (The sample proportion is often represented by p̂.)

7.2 The Sampling Distribution of the Sample Mean

6. Suppose that 16 observations are randomly selected from a normally distributed

population where µ = 10 and σ 2 = 625.

(a) What is the mean of the sampling distribution of X?

¯
(b) What is the standard deviation of the sampling distribution of X?
¯
(c) What is the distribution of X?
¯
(d) What is ¯
P (X > 30)
(e) What is P (0 < X¯ < 20)
1
While the probability of winning changes a tiny amount from cup to cup, as cups are revealed as winners
or non-winners, there are millions of cups so the changes in probability are minuscule. For the purposes
of these questions, assume that the cups can be considered independent.
Balka ISE 1.09 7.3. THE CENTRAL LIMIT THEOREM 97

7. Suppose that we intend to sample 400 observations from a large, normally dis-
tributed population. Unknown to us, the population mean and standard deviation
are both equal to 20. One of the plots in Figure 7.1 is the sampling distribution of
the sample mean in this scenario. Which one?

Figure 7.1: Which plot represents the sampling distribution of the sample mean?

7.3 The Central Limit Theorem

8. What does the central limit theorem tells us about the sampling distribution of the
sample mean? Why is this important in the world of statistics?

9. Consider the statement: The sampling distribution of µ is normal, provided n is

large. Is this statement true or false? Briefy justify your response.

10. We know that the sampling distribution of X ¯ has a mean of µ and a standard
deviation of √n . Which of the following statements are true?
σ

(a) The standard deviation of the sampling distribution of the sample mean de-
pends on the value of µ.
(b) We cannot possibly determine any characteristics of a statistic’s sampling dis-
tribution without repeatedly sampling from the population.
(c) The sampling distribution of X ¯ is always at least approximately normal for
large sample sizes, and is sometimes approximately normal for small sample
sizes.
(d) If the sample size is quadrupled, then the standard deviation of the sampling
distribution of the sample mean decreases by a factor of 2.
Balka ISE 1.09 7.4. SOME TERMINOLOGY REGARDING SAMPLING DISTRIBUTIONS 98

7.4 Some Terminology Regarding Sampling Distributions

7.4.1 Standard Errors

11. What is meant by the term standard error of a statistic? Does the term standard
error of a parameter have a similar meaning?

12. Suppose we randomly sample 6 values from a normally distributed population and
fnd that x̄ = 662.2 and s = 13.7. What is the standard error of the sample mean?

13. All else being equal, what is the e˙ect of the following on SE(X
¯ )?

(a) An increase in the sample size.

(b) An increase in the sample variance.
(c) An increase in the sample mean.

7.4.2 Unbiased Estimators

14. What is meant by the term unbiased estimator ?

15. Is the sample mean an unbiased estimator of the population mean?

7.5 Chapter Exercises

7.5.1 Basic Calculations

7.5.2 Concepts

16. In this chapter we have learned that when we randomly sample n observations from
a large population that has a mean of µ and a standard deviation of σ, the sampling
distribution of X
¯ has a mean of µ and a standard deviation of σ ¯ = √σ .
X n

Which of the following statements are true?

(a) The letter n represents the number of samples that are drawn in repeated
sampling, and not the number of observations in each sample.
(b) In practice, we usually take a large number of samples so that we can accurately
estimate the sampling distribution of X.¯
Balka ISE 1.09 7.5. CHAPTER EXERCISES 99

(c) The standard deviation of the population from which we are sampling de-
creases as the sample size increases.
(d) The standard deviation of the sampling distribution of X
¯ decreases as the
sample size increases.

17. Let X represent the number on the top face when an ordinary six-sided die is rolled
once, and let X¯ represent the average of the two numbers that come up when an
ordinary six-sided die is rolled twice.

(a) What is the probability distribution of X?

(b) What is E(X)?
(c) What is σ 2 ?
(d) What is the probability distribution of X?
¯
(e) What is the mean of the sampling distribution of X?¯ (There is a quick way
to fnd this and a slower way. Find the value using both methods.)
(f) What is the standard deviation of the sampling distribution of X?
¯ (There is a
quick way to fnd this and a slower way. Find the value using both methods.)

18. In Question 17, X represented the number that came up when an ordinary six-sided
die is rolled once, and it was found that E(X) = 3.5 and V ar(X) = 3512 . Let X
¯
represent the mean of the 2000 numbers that come up when the die is rolled 2000
times.

(a) What is the mean of X? ¯

(b) What is the standard deviation of X?
¯
(c) What does the central limit theorem tell us about the distribution of X?
¯
(d) Use the central limit theorem to approximate P (X
¯ < 3.5190).

19. Test your conceptual understanding: Which of the following statements are true?
(You should be able to explain why a statement is true or why a statement is false.)

(a) If we are sampling from a normally distributed population, then the sampling
distribution of X
¯ is normal for any sample size.
(b) The standard deviation of the sampling distribution of the sample mean in-
creases as the sample size increases.
(c) The central limit theorem states that the sample mean equals the population
mean, as long as the sample size is large.
(d) If the sampling distribution of X
¯ is approximately normal, then the population
from which we are sampling must be approximately normal.
(e) All else being equal, the mean of the sampling distribution of X¯ decreases as
the sample size increases.

20. Test your conceptual understanding: Which of the following statements are true?
(You should be able to explain why a statement is true or why a statement is false.)
Balka ISE 1.09 7.5. CHAPTER EXERCISES 100

(a) If we draw a very large sample from any population and plot a histogram of
the observations, the shape of the histogram will be approximately normal.
(b) In practice, we usually know the true standard deviation of the sampling dis-
tribution of X¯.
(c) In practice, we usually know the true value of µ.
(d) The sample mean is an unbiased estimator of the population mean.
(e) If we were to repeatedly sample from a population, then the distribution of the
sample mean would become approximately normal as the number of samples
increases, as long as the sample size of each sample stays constant.

21. Test your conceptual understanding: Which of the following statements are true?
(You should be able to explain why a statement is true or why a statement is false.)

(a) Statistics have sampling distributions.

(b) The value of a parameter does not vary from sample to sample.
(c) The value of a statistic does not vary from sample to sample.
(d) All else being equal, the standard deviation of the sampling distribution of the
sample mean will be smaller for n = 10 than for n = 40.
(e) The sampling distribution of µ is usually approximately normal for n > 30.

7.5.3 Applications

22. A study2 found that the amount of ribofavin (vitamin B2) in large eggs is approx-
imately normally distributed with a mean of 0.25 mg and a standard deviation of
0.02 mg.

(a) What proportion of large eggs have less than 0.20 mg of ribofavin?
(b) What is the distribution of the mean amount of ribofavin in 30 large eggs?
(c) What is the probability that the average amount of ribofavin in 30 large eggs
lies between 0.240 and 0.260 mg?
(d) What is the probability that the average amount of ribofavin in 30 large eggs
lies between 0.250 and 0.255 mg?
(e) What is the 20th percentile of the distribution of the average amount of ri-
bofavin in 30 large eggs?

23. The amount of protein in a popular breakfast sandwich from a large fast chain
is approximately normally distributed with a mean of 17.0 grams and a standard
deviation of 0.80 grams. (See footnote 2.)
2
The mean and standard deviation in this question are estimates based on information from the USDA
National Nutrient Database. Assume that they are the true values for the purposes of this question.
Balka ISE 1.09 7.5. CHAPTER EXERCISES 101

(a) If one of these breakfast sandwiches is randomly selected, what is the proba-
bility it contains less than 16.0 grams of protein?
(b) What is the distribution of the total amount of protein in 2 randomly selected
breakfast sandwiches of this type?
(c) What is the distribution of the mean amount of protein in 2 randomly selected
breakfast sandwiches of this type?
(d) If 2 breakfast sandwiches of this type are randomly selected, what is the prob-
ability the total amount of protein is less than 32.0 grams?
(e) If 2 breakfast sandwiches of this type are randomly selected, what is the prob-
ability the mean amount of protein is at least 16.5 grams?
(f) What is the 97.5th percentile of the distribution of the average amount of
protein in 2 randomly selected sandwiches of this type?

24. A study3 found that the length of the right ear in Caucasian Italian men between
the ages of 18 and 30 is approximately normally distributed with a mean of 62 mm
and a standard deviation of 4 mm.

(a) If one Caucasian Italian man in this age group is randomly selected, what is
the probability the length of his right ear is at least 60.0 mm?
(b) What is the 90th percentile of the length of the right ear in Caucasian Italian
men of this age group?
(c) If 100 Caucasian Italian men are randomly selected, what is the distribution
of the average length of their right ears?
(d) If 100 Caucasian Italian men are randomly selected, what is the probability
that the average length of their right ears is greater than 61.8 mm?
(e) What is the 10th percentile of the distribution of the average right ear length
in 100 randomly selected Caucasian Italian men?

25. A study4 found that that carapace lengths of adult male crayfsh (Astacus lepto-
dactylus) in a lake in Turkey are approximately normally distributed with a mean
of 48.0 mm and a standard deviation of 2.5 mm.

(a) If a single adult male crayfsh from this lake is randomly selected, what is the
probability its carapace length is less than 50.0 mm?
(b) What is the distribution of the mean carapace length of 10 randomly selected
adult male crayfsh from this lake?
(c) If 10 adult male crayfsh are randomly selected from this lake, what is the
probability their average carapace length is less than 50.0 mm?
(d) If 10 adult male crayfsh are randomly selected from this lake, what is the
3
Sforza et al. (2009). The mean and standard deviation given in this question are estimates based on
this article, but for the purposes of this question assume that µ = 62 and σ = 4.
4
?. The mean and standard deviation given in this question are estimates based on this article, but for
the purposes of this question assume that µ = 48.0 and σ = 2.5.
Balka ISE 1.09 7.5. CHAPTER EXERCISES 102

probability their average carapace length lies between 46.0 mm and 50.0 mm?
(e) What is the 95th percentile of the distribution of the average carapace length
of 10 randomly selected adult male crayfsh from this lake?

26. A study5 found that birth weights of female African elephants born in captivity are
approximately normally distributed with a mean of 95.1 kg and a standard deviation
of 13.7 kg. (The following questions are all in reference to African elephants born in
captivity. The distribution of weights of elephants born in the wild is likely di˙erent
from that of those born in captivity.)

(a) What is the probability that a randomly selected newborn female African
elephant weighs more than 100.0 kg?
(b) What is the sampling distribution of the mean weight of 3 randomly selected
newborn African elephants?
(c) What is the probability that the mean weight of 3 randomly selected newborn
female African elephants is more than 100.0 kg?
(d) What is the probability that the mean weight of 10 randomly selected newborn
female African elephants lies between 98.0 and 102.0 kg?
(e) What is the 30th percentile of birth weight of female African elephants?
(f) What is the 30th percentile of the mean birth weight of 10 randomly selected
newborn female African elephants?

27. Many studies have investigated the weight of a human organs, often basing the
fndings on post-mortem results from accidental deaths. In one study6 , it was found
that the human heart in adult Caucasian males had a mean of 365.0 grams and a
standard deviation of 71.0 grams. For the purposes of this question, assume that
the heart weight of adult Caucasian males is approximately normally distributed
with a mean of 365.0 grams and a standard deviation of 71.0 grams.

(a) What is the value of the 90th percentile of heart weight for adult Caucasian
males?
(b) Suppose 9 adult Caucasian males are randomly selected. What is the sampling
distribution of their mean heart weight?
(c) If 9 adult Caucasian males are randomly selected, what is the probability the
average weight of their hearts is greater than 350.0 grams?
(d) If 9 adult Caucasian males are randomly selected, what is the probability the
average weight of their hearts is between 340 and 380.0 grams?
(e) If 9 adult Caucasian males are randomly selected, what is the 90th percentile
of the sampling distribution of their mean heart weight?
5
Dale (2010). The mean and standard deviation given in this question are estimates based on this article,
but for the purposes of this question assume that µ = 95.1 and σ = 13.7.
6
Lorin de la Grandmaison et al. (2001). Organ weight in 684 adult autopsies: new tables for a Caucasoid
population. Forensic Science International, 119:149–154.
Balka ISE 1.09 7.5. CHAPTER EXERCISES 103

(f) If 4 adult Caucasian males are randomly selected, what is the probability that
the sum of the weights of their hearts is greater than 1400 grams?

28. A study found that heights of Canadian females aged 20 to 39 are approximately
normally distributed with a mean of 163.3 cm and a standard deviation of 6.4 cm.7

The heights of adult females in Canada are approximately normally distributed

with a mean of 164.0 cm, and a standard deviation of 6.9 cm.

(a) What is the probability that a randomly selected Canadian female in this age
group is at least 170.0 cm tall?
(b) If 100 Canadian females in this age group are randomly selected, what is the
sampling distribution of their average height?
(c) If 100 Canadian females in this age group are randomly selected, what is the
probability their average height is less than 165.0 cm?
(d) If 100 Canadian females in this age group are randomly selected, what is the
probability their average height is between 163.0 and 165.0 cm?
(e) For a random sample of 100 Canadian females in this age group, what are
the 2.5th and 97.5th percentiles of the sampling distribution of their average
height?

7.5.4 Extra Practice Questions

29. Suppose that marks on a test in a very large statistics course are approximately
normally distributed with a mean of 71 and a variance of 100. Suppose that we are
about to randomly sample 5 of the students that wrote the test.

(a) What is the mean of the sampling distribution of the mean score of the 5
students?
(b) What is the standard deviation of the sampling distribution of the mean score
of the 5 students?
(c) What is the distribution of the sampling distribution of the mean score of the
5 students?
(d) If a single student is randomly selected, what is the probability they scored
higher than 75.0?
(e) If 5 students are randomly selected, what is the probability their mean score
is greater than 75.0?
(f) If 5 students are randomly selected, what is the probability they all scored
greater than 75.0?
7
The mean and standard deviation are estimates based on information from the Canadian Health Mea-
sures Survey (Statistics Canada, 2013). Assume these are the true values for the purposes of this
question.
Balka ISE 1.09 7.5. CHAPTER EXERCISES 104

30. A type of steel shaft is produced by a process that results in the shaft diameters
being approximately normally distributed with a mean of 4.10 cm and a standard
deviation of 0.05 cm.

(a) What is the probability that a randomly selected shaft has a diameter less
than 4.23 cm?
(b) If 16 shafts are randomly selected, what is the sampling distribution of their
mean diameter?
(c) If 100 shafts are randomly selected, what is the sampling distribution of their
mean diameter?
(d) If 5 shafts are randomly selected, what is the probability their mean diameter
lies between 4.05 and 4.15 cm?
(e) If 5 shafts are randomly selected, what is the probability their mean diameter
is at least 4.15 cm?
(f) What is the 90th percentile of the distribution of the mean diameter of 20
randomly selected shafts?

31. A supplier of pure oxygen claims that their containers of pure oxygen have a mean
purity of 99.500%, with a standard deviation of 0.048%. In your frst dealing with
this company, your frst container has a purity level of 99.43%. Assume the purity
levels in these containers are approximately normally distributed.

(a) Assuming the company’s claim is true, what is the probability that a randomly
selected container has a purity level less than 99.43%?
(b) Assuming the company’s claim is true, what is the probability that the mean
of 10 randomly selected containers is less than 99.45%?
(c) Assuming the company’s claim is true, what is the probability that the mean
of 10 randomly selected containers lies between 99.45% and 99.55%?
(d) Suppose in your frst 25 containers purchased from this company, you fnd a
mean oxygen purity of 99.00%. Does this give strong evidence against the
company’s claimed purity level?
Chapter 8

Confdence Intervals

J.B.’s strongly suggested exercises: 1, 2, 4, 5, 7, 8, 9, 10, 11, 13, 14, 24, 26, 27, 29, 31, 32

8.1 Introduction

8.2 Interval Estimation of µ when σ is Known

1. Suppose we are sampling from a normally distributed population and we wish to

calculate a confdence interval for the population mean µ. Find the appropriate z
value and margin of error in the following situations.

(a) n = 22, σ = 5, 95% confdence level.

(b) n = 10, σ = 10, 90% confdence level.
(c) n = 18, σ = 15, 99% confdence level.
(d) n = 18, σ = 15, 63.2% confdence level.
(e) n = 50, σ = 35, 21.2% confdence level.

2. Suppose we draw a random sample of n observations from a normally distributed

population, and we wish to construct a confdence interval for the population mean
µ. Suppose that the population standard deviation σ is known. Give the confdence
level for each of the following intervals.

(a) X
¯ ± 1.21 √σ
n
(b) X
¯ ± 1.35 √σ
n
(c) X
¯ ± 1.57 √σ
n

105
Balka ISE 1.09 8.2. INTERVAL ESTIMATION OF µ WHEN σ IS KNOWN 106

(d) X
¯ ± 2.45 √σ
n
(e) X
¯ ± 3.15 √σ
n

3. Which of the following statements are true? (You should be able to explain why a
statement is true or why a statement is false.)

(a) In statistical inference, we often construct confdence intervals for statistics.

(b) A confdence interval is a range of plausible values for a parameter.
(c) The confdence level of an interval is determined by sample data.
(d) The point estimate of µ lies at the midpoint of the confdence interval for µ.
(e) The confdence interval procedures of this section assume a normally dis-
tributed population, but this assumption becomes less important as the sample
size increases.

8.2.1 Interpretation of the Interval

4. Researchers are interested in estimating the population mean body mass index
(BMI) for female students at a university. (A person’s BMI is their weight in
kilograms divided by the square of their height in metres.) Measurements on 200
female student volunteers at this university resulted in an average BMI of 25.7
kg/m2 . (For the purposes of this question, assume that the population standard
deviation is 6.0 kg/m2 . (This value is likely not far from reality.) Also assume that
the volunteers can be thought of as a random sample of female students from this
university. This assumption may or may not be reasonable, depending on how the
volunteers were obtained.)

(a) What is a 95% confdence interval for the population mean?

(b) What is an appropriate interpretation of this interval?
(c) What is a 99% confdence interval for the population mean?
(d) What is an appropriate interpretation of this interval?

5. A study by Sforza et al. (2009) investigated physical characteristics of the ears of

Caucasian Italians. In one part of the study, a sample of 66 females between the
ages of 18 and 30 years old revealed a mean right ear length of 61.9 mm with a
standard deviation of 4.1 mm. The resulting 95% confdence interval for µ was
found to be (60.91, 62.89). Assuming the 66 females in the sample can be thought
of as a random sample of Caucasian Italian females between the ages of 18 and 30,
give an appropriate interpretation of this confdence interval in the context of the
situation at hand.

6. Suppose that at a certain university, a random sample of 50 graduating students

yielded an average student loan debt of $28,500, with an associated 95% confdence
Balka ISE 1.09 8.3. CONFIDENCE INTERVALS FOR µ WHEN σ IS UNKNOWN 107

interval for µ of (26,500, 30,500). Which, if any, of the following options are correct
interpretations of this interval?

(a) 95% of graduating students at this university have student loan debt between
$26,500 and $30,500.
(b) We can be 95% confdent that the average student loan debt of the 50 students
in the sample lies between $26,500 and $30,500.
(c) We can be 95% confdent that the true mean student loan debt of graduating
students at this university lies between $26,500 and $30,500.
(d) In repeated sampling, 95% of the 95% confdence intervals would contain
$28,500.

8.2.2 What Factors A˙ect the Margin of Error?

7. All else being equal, what is the e˙ect of the following on the margin of error of a
confdence interval for µ?

(a) An increase in the sample size.

(b) An increase in the variance.
(c) An increase in the sample mean.
(d) An increase in the confdence level.

8.2.3 An Example

8.3 Confdence Intervals for µ When σ is Unknown

8.3.1 Introduction

8. Suppose we are sampling from a normally distributed population and we wish to

calculate a confdence interval for the population mean µ. Find the appropriate t
value, standard error of X,
¯ and margin of error in the following scenarios.

(a) n = 14, s = 10, and an 80% confdence level.

(b) n = 22, s = 10 and an 80% confdence level.
(c) n = 8, s = 2, 500 and a 90% confdence level.
(d) n = 8, s = 2, 500 and a 95% confdence level.
(e) n = 8, s = 3.2, and a 99% confdence level.

9. Test your conceptual understanding: Which of the following statements are true?
(You should be able to explain why a statement is true or why a statement is false.)
Balka ISE 1.09 8.3. CONFIDENCE INTERVALS FOR µ WHEN σ IS UNKNOWN 108

(a) The variance of the t distribution is greater than the variance of the standard
normal distribution.
(b) The mean of the t distribution is equal to the mean of the standard normal
distribution.
(c) The t distribution has more area in the tails and a lower peak than the standard
normal distribution.
(d) As the degrees of freedom increase, the t distribution tends toward the standard
normal distribution.
(e) The t distribution is equivalent to the standard normal distribution if the
degrees of freedom are at least 30.

10. Suppose that we should be using tα/2 in the formula to fnd a confdence interval
for µ, but we mistakenly use zα/2 in its place. Would the interval found using zα/2
be wider or narrower than if we had used the appropriate tα/2 value?

8.3.2 Examples

11. As part of an investigation by Fink et al. (2007), 32 male student volunteers at a

university in Germany had their hand-grip strength measured (in kilograms force
(kgf)). The results are illustrated in Figure 8.1. The sample mean and standard
deviation were found to be 50.91 kgf and 7.74 kgf, respectively.
Hand Grip Strength (kgf)

60

60
Sample Quantiles
55

55
50

50
45

45
40

40
35

35

-2 -1 0 1 2
Theoretical Quantiles

Figure 8.1: Plots of the grip strengths of 32 male student volunteers.

For the purposes of this question, assume that the volunteers can be thought of as
a random sample of male students at this German university. (This may or may
not be a reasonable assumption, depending on the specifcs of how the volunteers
were recruited.)

(a) Do these plots show a violation of the normality assumption?

(b) What is the point estimate of the population mean µ? What is the value of
the standard error of the sample mean?
(c) What is a 95% confdence interval for the population mean µ?
Balka ISE 1.09 8.4. DETERMINING THE MINIMUM SAMPLE SIZE N 109

(d) Give an appropriate interpretation of the 95% confdence interval in the context
of the situation at hand. To what population do your conclusions apply?
Comment on any biases that might be present.

8.3.3 Assumptions of the One-Sample t Procedures

12. Suppose we intend to use the t procedure to construct a confdence interval for the
population mean µ. If the population is not normally distributed, what are the
implications?

13. What types of violation of the normality assumption are very problematic for the
t procedure? What types of violation of the normality assumption are not a big
problem?

8.4 Determining the Minimum Sample Size n

14. Suppose we are about to sample from a normally distributed population, and we
wish to estimate the population mean µ. Calculate the minimum sample size re-
quired in the following scenarios.

(a) σ = 10, and we wish to estimate µ to within 5 with 90% confdence.

(b) σ = 10, and we wish to estimate µ to within 5 with 95% confdence.
(c) σ = 30, and we wish to estimate µ to within 5 with 95% confdence.
(d) σ = 10, and we wish to estimate µ to within 0.1 with 99% confdence.
(e) σ 2 = 16, and we wish to estimate µ to within 1 with 95% confdence.

8.5 Chapter Exercises

8.5.1 Basic Calculations

15. Suppose we are about to randomly sample 16 values from a population where µ = 5
and σ = 10. What is the value of σX̄ ?

16. A random sample of 16 independent observations had a mean of 20.0 and a standard
deviation of 2.0. What is the value of SE(X
¯ )?

17. Suppose the following sets of values represent samples drawn from normally dis-
tributed populations. For each scenario, calculate a 95% confdence interval for the
population mean µ.
Balka ISE 1.09 8.5. CHAPTER EXERCISES 110

(a) 3, 5, 12.
(b) −4, 8, 23.
(c) 2012, 2010, 2054, 2032.
(d) −52, −48, −48, −46, −42.

8.5.2 Concepts

18. A person is interested in estimating the average weight (in kg) of adult males of a
certain breed of dog. They draw a sample of adult males of this species, measure
their weights, and fnd that the 95% confdence interval for the true mean weight
is (12.2, 14.9). Of the following options, which one is the best interpretation of this
interval? (Assume that the sample can be thought of as a simple random sample
of adult males of this breed.)

(a) In repeated sampling, 95% of the time µ will lie between 12.2 kg and 14.9 kg.
(b) In repeated sampling, 95% of the sample mean weights will lie between 12.2
kg and 14.9 kg.
(c) In repeated sampling, 95% of the 95% confdence intervals will be (12.2, 14.9).
(d) We can be 95% confdent that the population mean weight of all adult males
of this breed lies between 12.2 kg and 14.9 kg.
(e) 95% of adult males of this breed weigh between 12.2 kg and 14.9 kg.

19. Suppose we repeatedly draw random and independent samples from a normally
distributed population and for each sample we calculate a 95% confdence interval
for the population mean µ. In 100 of such intervals:

(a) What is the distribution of the number of intervals that contain µ?

(b) What is the probability that at least one interval does not contain µ?

20. Suppose that a researcher is interested in estimating the average survival time of
rats after they have been infected with a certain type of virus. The researcher infects
100 rats with the virus and then measures the time until death. The average time
until death is found to be 1.4 months. (For the purposes of this question, assume
that σ is known to be 1.0 months.)

(a) Survival times are often skewed to the right. If that is the case here, would
inference methods based on the normal distribution be inappropriate?
(b) What is the the value of the standard deviation of the sampling distribution
of the sample mean?
(c) What is a 92% confdence interval for the mean µ?
(d) What is the true value of µ?
(e) What does µ represent in this situation?
Balka ISE 1.09 8.5. CHAPTER EXERCISES 111

21. True or false? tα/2 tends toward zα/2 as the degrees of freedom increase.

22. (Warning: This is a very challenging question!)

In a (not very helpful) video on YouTube, a statistics instructor at a university
talks about interpreting a confdence interval for the population mean, and he
states that if we were to repeatedly sample from the population, 95% of the sample
means would fall within the 95% confdence interval. This is simply untrue. A
correct interpretation is that in repeated sampling, 95% of the calculated intervals
would capture the true value of the population mean µ.

This question was motivated by that YouTube video. Suppose that we are about
to draw a simple random sample of n observations from a normally distributed
population, then calculate a 95% confdence interval for µ. (Suppose that σ is
¯ ± z.025 √σ .) If we were
known, so we will calculate the interval using the formula: X n
to draw another, independent sample of the same size from this population, what
is the probability that the sample mean of this second sample falls within the 95%
confdence interval for µ found in the frst sample? (The probability is not 0.95, as
you may guess from the fact that this question is being asked.)

23. (Warning: This question is not very complicated, but most students fnd it concep-
tually diÿcult. If you don’t understand it, don’t spend too much time on it.)
A researcher wants to calculate a 95% confdence interval for a parameter θ. The
statistic θ̂ is a point estimator of θ, and is more complicated than any estimator
that we have discussed. The researcher draws a random sample (n = 300), and
fnds that θ̂ = 2.00 for her sample. Suppose it is known that for samples of this
size, the sampling distribution of θ̂ is normal with a mean of θ and a standard
deviation of 3.00. What is a 95% confdence interval for θ?

24. Test your conceptual understanding: Which of the following statements are true?
(You should be able to explain why a statement is true or why a statement is false.)

(a) The standard error of a statistic decreases as the sample size decreases.
(b) The one-sample t procedures are robust to violations of the normality assump-
tion.
(c) The t procedures do not perform well when there is strong skewness or outliers
in the data, especially for small sample sizes.
(d) The width of a confdence interval for µ decreases as the sample size decreases.
(e) The width of a confdence interval for µ decreases as the variance increases.

25. Test your conceptual understanding: Which of the following statements are true?
(You should be able to explain why a statement is true or why a statement is false.)

(a) In practical problems, σ is usually known and thus we usually use z procedures
instead of t procedures.
Balka ISE 1.09 8.5. CHAPTER EXERCISES 112

(b) Standard errors are the estimated standard deviations of parameters in re-
peated sampling.
(c) If the assumptions of a procedure used to calculate a 90% confdence interval
are violated, the true confdence level of the interval may be very di˙erent
from 90%.
(d) All else being equal, the width of a confdence interval decreases as the conf-
dence level increases.
(e) All else being equal, the margin of error decreases as the confdence level
increases.
(f) All else being equal, the standard error of the sample mean decreases as the
confdence level of the interval increases.
(g) All else being equal, the confdence level of an interval increases as the sample
size increases.

8.5.3 Applications

Weight
26. Body mass index (BMI = Height2
, where weight is in kg and height is in m) is
a measure of body shape. A study1 investigated BMI in Texas youth. Student
participants were recruited from 6 schools in a school district in a rural area of
North Central Texas (about 100 km from Fort Worth). Students participated only
if their parents provided informed consent. Students who participated were given
a free t-shirt.

Here will look only at the BMI data for the 94 14 year-old boys in the study.
Figure 8.2 illustrates the BMI values for the 94 boys. For the boys in this sample,
the mean BMI is 23.3, with a standard deviation of 5.4.

(a) Do these plots show a violation of the normality assumption? If there is a

violation, is it a serious problem in this situation? (Can we still use the t
procedures?)
(b) What is the point estimate of the population mean µ? What is the value of
the standard error of the sample mean?
(c) What is a 95% confdence interval for the population mean µ?
(d) Give an appropriate interpretation of the 95% confdence interval in the context
of the situation at hand. To what population do your conclusions apply?
Comment on any biases that might be present.

27. Franklin et al. (2012)2 investigated various aspects of tandem running in ants. Tan-
1
Duran et al. (2013). Growth and weight status of rural texas school youth. American Journal of Human
Biology, 25:71–77. The data used here is simulated data with the same summary statistics as given in
their Table 1.
2
Franklin et al. (2012). Do ants need to be old and experienced to teach? The Journal of Experimental
Balka ISE 1.09 8.5. CHAPTER EXERCISES 113

40

40
35

35
Sample Quantiles
30

30
BMI
25

25
20

20
15

15
-2 -1 0 1 2
Theoretical Quantiles

(a) Boxplot. (b) Normal quantile-quantile plot.

Figure 8.2: BMI for 94 boys in rural Texas.

dem running is a form of recruitment in which one ant with knowledge of the loca-
tion of a food source or new nest site leads another ant to that location. (Optional:
Watch an example of tandem running here: (1:53) (http://www.youtube.com/watch?v=X2C7Sy2oPik))
Tandem running can be thought of as a form of teaching and learning.

The study investigated various factors associated with tandem running, and inves-
tigated whether the age of the ant and the experience of the ant had any e˙ect on
tandem running. Ants (Temnothorax albipennis) were categorized into 4 categories:
Young and inexperienced (YI), young and experienced (YE), old and inexperienced
(OI), old and experienced (OE). One aspect of the study investigated the speed of
tandem running for 51 tandem runs in which a YE ant was leading a YI ant. The
mean speed (mm/s) of the run was recorded. Figure 8.3 illustrates the speed of the
runs. For these 51 runs, the sample mean was 1.65 mm/s and the sample standard
deviation was 0.62 mm/s.
2.5

2.5
Sample Quantiles
Speed (mm/s)
2.0

2.0
1.5

1.5
1.0

1.0
0.5

0.5

-2 -1 0 1 2
Theoretical Quantiles

(a) Boxplot. (b) Normal quantile-quantile plot.

Figure 8.3: Speed of 51 tandem runs.

(a) Do these plots show a violation of the normality assumption? If there is a

Biology, 215:1287–1292 The data used here is simulated data based on their Figure 2, with similar results
and conclusions.
Balka ISE 1.09 8.5. CHAPTER EXERCISES 114

violation, is it a serious problem in this situation? (Can we still use the t

procedures?)
(b) What is the point estimate of the population mean µ? What is the value of
the standard error of the sample mean?
(c) What is a 95% confdence interval for the population mean µ?
(d) Give an appropriate interpretation of the 95% confdence interval, in the con-
text of the situation at hand. To what population do your conclusions apply?
Comment on any biases that might be present.

28. The Athens Collection is a collection of 225 skeletal remains, housed at the Univer-
sity of Athens in Greece. The remains were collected from cemeteries near Athens,
from people who lived mainly in the second half of the 20th century. A study3 in-
vestigated the diameter of molar teeth of individuals in this collection. Of 24 upper
M1 molars from adult males, the mean crown diameter was found to be 10.38 mm
with a standard deviation of 0.63 mm.

(a) Suppose we wish to construct a 95% confdence interval for the mean crown
diameter of upper M1 molars in adult males in Greece. What assumptions
are necessary for the t confdence interval procedure to be valid? Are the
assumptions likely to be satisfed in this scenario?
(b) What is the point estimate of the population mean µ? What is the value of
the standard error of the sample mean?
(c) Assuming normality, construct a 95% confdence interval for the population
mean.
(d) Give an appropriate interpretation of the 95% confdence interval, in the con-
text of the situation at hand. To what population do your conclusions apply?
Comment on any biases that might be present.

29. Larviposition is the laying of larvae (the eggs hatch inside the female), and it occurs
in a small number of insect species. Cook et al. (2012) investigated various aspects
of larviposition of the blowfy Calliphora varifrons. In one aspect of the study,
female fecundity was investigated by trapping female C. varifrons blowfies in a
feld, dissecting them, and counting the number of live larvae they had inside. Of
the 49 females that carried live larvae, the mean number of larvae was 33.4, with a
standard deviation of 7.0.

(a) Suppose we wish to construct a 95% confdence interval for the mean number
of live larvae carried by female C. varifrons blowfies in this feld. What
assumptions are necessary for the t confdence interval procedure to be valid?
Are the assumptions likely to be satisfed in this scenario?
(b) Although the full data was not reported in the original journal article, the
3
Zorba et al. (2013). Evaluation of the accuracy of di˙erent molar teeth measurements in assessing sex.
Forensic Science, Medicine, and Pathology, 9:13–23.
Balka ISE 1.09 8.5. CHAPTER EXERCISES 115

authors did report that the number of live larvae varied between 20 and 44 for
the 49 females in the study. What does this tell us about the validity of the t
procedure in this scenario?
(c) What is the point estimate of the population mean µ? What is the value of
the standard error of the sample mean?
(d) Assuming normality, construct a 95% confdence interval for the population
mean.
(e) Give an appropriate interpretation of the 95% confdence interval, in the con-
text of the situation at hand. To what population do your conclusions apply?
Comment on any biases that might be present.

30. In another aspect of the study frst discussed in Question 29, Cook et al. (2012)
presented 42 female C. varifrons blowfies with a cube of fresh beef liver, and
measured the rate of larviposition. The rate of deposition (number of larvae laid
per second) had an average of 0.46, with a standard deviation of 0.32.

(a) Suppose we wish to construct a 95% confdence interval for the true mean rate
of larvae deposition for female C. varifrons blowfies under the conditions of
this experiment. What assumptions are necessary for the t confdence inter-
val procedure to be valid? Are the assumptions likely to be satisfed in this
scenario?
(b) What is the point estimate of the population mean µ? What is the value of
the standard error of the sample mean?
(c) Assuming normality, construct a 95% confdence interval for the population
mean.
(d) Give an appropriate interpretation of the 95% confdence interval, in the con-
text of the situation at hand. To what population do your conclusions apply?
Comment on any biases that might be present.

31. The nutrition information published by a popular fast-food chain claims that, in
their U.S. locations, a serving of hash browns contains 310 mg of sodium. A sample
of 6 servings of these hash browns resulted in a mean sodium content of 324.8 mg
and a standard deviation of 40.0 mg.4

(a) Suppose we wish to construct a 95% confdence interval for the true mean
sodium content in servings of hash browns from U.S. locations of this fast-food
chain. What assumptions are necessary for the t confdence interval procedure
to be valid? Are these assumptions likely to be satisfed in this scenario?
(b) What is the point estimate of the population mean µ? What is the value of
the standard error of the sample mean?
(c) The sample size is quite small here (n = 6), so the normality assumption is
very important. It is always a little sketchy using the t procedures for such a
4
These values are based on sample data found in the USDA National Nutrient Database.
Balka ISE 1.09 8.5. CHAPTER EXERCISES 116

small sample size, but none of the observations (not shown here) were outliers,
and there were no obvious problems with the data, so the use of the t procedure
might be reasonable. Assuming normality, and assuming that the sample of
hash browns can be thought of as a random sample of hash browns from U.S.
locations of this chain, construct a 95% confdence interval for the population
mean.
(d) Give an appropriate interpretation of the 95% confdence interval, in the con-
text of the situation at hand. To what population do your conclusions apply?
Comment on any biases that might be present.
(e) The chain’s claimed mean sodium content is 310 mg. Based on our analysis, is
this claimed value a plausible value of the population mean sodium content?

32. The ratio of the length of the index fnger to the length of the ring fnger is called
the 2D:4D ratio. This ratio is believed to be infuenced by fetal exposure to
testosterone, and it has been linked to characteristics such as aggression and to
diseases such as certain types of cancer. The ratio tends to be less than 1 in both
sexes (the index fnger tends to be shorter than the ring fnger), and the 2D:4D
ratio tends to be smaller in men than in women. The distribution of the 2D:4D
ratio di˙ers between groups of di˙erent ethnicities.

A study5 investigated the 2D:4D ratio in samples of women drawn from a sexual
health clinic in Manchester, UK. In one aspect of the study, it was found that 46
black women had a mean 2D:4D ratio (of the left hand) of 0.963 with a standard
deviation of 0.034, and the distribution was found to be “close to normal”.

(a) Suppose we wish to construct a 95% confdence interval for the true mean
2D:4D ratio for black women in the Manchester area. What assumptions
are necessary for the t confdence interval procedure to be valid? Are the
assumptions likely to be satisfed in this scenario?
(b) What is the point estimate of the population mean µ? What is the value of
the standard error of the sample mean?
(c) Construct a 95% confdence interval for the population mean.
(d) Give an appropriate interpretation of the 95% confdence interval, in the con-
text of the situation at hand. To what population do your conclusions apply?
Comment on any biases that might be present.

33. Leth (2009) investigated various aspects of homicides in Denmark. One aspect
of the study involved investigating characteristics of intimate partner homicide (a
person murdered their intimate partner). In 36 intimate partner homicides over
the years 1983–2007, the average age of the perpetrator was 43.8 years and the
standard deviation was 15.5 years.
5
Brabin et al. (2008). The second to fourth digit ratio (2d:4d) in women with and without human
papillomavirus and cervical dysplasia. American Journal of Human Biology, 20:337–341
Balka ISE 1.09 8.5. CHAPTER EXERCISES 117

(a) Suppose we wish to construct a 95% confdence interval for the true (theoret-
ical) mean age of the perpetrators of intimate partner homicides in Denmark
over these years. What assumptions are necessary for the t confdence inter-
val procedure to be valid? Are the assumptions likely to be satisfed in this
scenario?
(b) What is the point estimate of the population mean µ? What is the value of
the standard error of the sample mean?
(c) Assuming normality, construct a 95% confdence interval for the population
mean.
(d) Give an appropriate interpretation of the 95% confdence interval, in the con-
text of the situation at hand. To what population do your conclusions apply?
Comment on any biases that might be present.

8.5.4 Extra Practice Questions

34. A new airline is interested in estimating the mean fying time for 747 fights from
Toronto (YYZ) to Vancouver (YVR). A sample of 15 fights yielded a mean time of
295.6 minutes and a sample standard deviation of 10.2 minutes. The airline wishes
to use the data to construct a 95% confdence interval for the population mean fight
time. They (wisely) think that they should check the normality assumption before
using a procedure based on the assumption of a normally distributed population.
They plot a normal quantile-quantile plot of the 15 fight times, and the results are
illustrated in Figure 8.4.

Normal Q-Q Plot

310
305
300
Sample Quantiles

295
290
285
280

-1 0 1

Theoretical Quantiles

Figure 8.4: Normal QQ plot for the 15 fight times in Question 34.

(a) Does this plot give any indication that the t procedures should not be used
here?
Balka ISE 1.09 8.5. CHAPTER EXERCISES 118

(b) Assuming that the sample of fight times can be thought of as a simple random
sample from the population of fight times, and assuming that the fight times
are approximately normally distributed, calculate a 95% confdence interval
for the population mean fight time.
(c) Suppose that the interval in 34b was found to be (288.1, 296.1). This is not
the correct interval but assume that it is for the purposes of this question. Of
the following options, which one is the best interpretation of that interval?
i. 95% of YYZ–YVR fights take between 288.1 and 296.1 minutes.
ii. We can be 95% confdent that the population mean fight time from YYZ
to YVR for these types of fights lies between 288.1 and 296.1 minutes.
iii. We can be 95% confdent that the sample mean fight time of the 15 fights
lies between 288.1 and 296.1 minutes.
iv. A randomly selected YYZ–YVR fight time has a 95% chance of taking
between 288.1 and 296.1 minutes.
v. All of the above.

35. Suppose you wish to estimate the average arsenic level in the soil of playgrounds
of elementary schools in a very large school district. You randomly sample schools
from this district, fnd that the sample mean arsenic level is 14.60 ppm with a 95%
confdence interval for µ of (14.1, 15.1). Of the following options, which one is the
most appropriate interpretation of that interval?

(a) In repeated sampling, 95% of the sample means will fall between 14.1 and 15.1
ppm.
(b) We can be 95% confdent that the mean arsenic level of the schools in the
sample lies between 14.1 and 15.1 ppm.
(c) In repeated sampling, 95% of the time the confdence interval for µ will be
(14.1, 15.1).
(d) We can be 95% confdent that the population mean arsenic level of the soils
of schools in this district lies between 14.1 and 15.1 ppm.
(e) In repeated sampling, µ will change from sample to sample. 95% of the time
µ will lie between 14.1 and 15.1 ppm, and 5% of the time it will be outside of
this interval.

36. Executives at a large company wish to investigate the average wait time for phone
calls that are received at a call centre. The wait times (in minutes) for a random
sample of 600 phone calls to the call centre are recorded, entered into the statistical
software R, and the following output is found.
Balka ISE 1.09 8.5. CHAPTER EXERCISES 119

One Sample t-test

data: calltime
t = 50.0812, df = 599, p-value < 2.2e-16
alternative hypothesis: true mean is not equal to 0
95 percent confidence interval:
2.884044 3.119472

Give an appropriate interpretation of the confdence interval found in the output.

37. A car manufacturer is investigating the fuel consumption of two types of engine
design. As part of the study, 8 engines of Type A are placed into a certain type
of car and driven under real-world conditions. The cars are driven under similar
conditions and the fuel consumption (litres/100 km) is recorded. The eight cars had
a mean fuel consumption of 6.8 l/100k with a standard deviation of 0.32 l/100k.

(a) If we had access to the data, how should we go about investigating whether
the fuel consumption values are approximately normally distributed?
(b) Suppose that it is reasonable to assume that the fuel consumption values are
approximately normally distributed. What is a 95% confdence interval for the
mean fuel consumption for this type of engine?
(c) Give an appropriate interpretation of the 95% interval.

38. As part of a study into how many facial tissues should be put into a standard
size box, a facial tissue company wants to investigate how many facial tissues a
cold su˙erer will use per day on average. They randomly sample 26 cold su˙erers,
watch them for a day, and fnd that they use an average of 29.3 facial tissues.
The sample standard deviation was found to be 11.9. Assume the population is
normally distributed in this situation (but keep in mind that this is a bit of a
sketchy assumption).

(a) What is the point estimate of the population mean?

(b) What is the estimated value of the standard deviation of the sampling distri-
bution of the sample mean? In other words, what is the standard error of the
sample mean?
(c) What is the appropriate 95% confdence interval for µ, the population mean
number of facial tissues used per day?
(d) What is the 90% margin of error?
(e) What is the appropriate 90% confdence interval?
(f) What is an appropriate interpretation of the 90% interval?

39. Food scientists developed a new type of hydrogenated vegetable oil, and they are
investigating several properties of this oil. As part of the investigation, they want
to know the average melting temperature of the new oil. A random sample of 10
batches of this oil are heated until they melt, and the melting temperature is then
Balka ISE 1.09 8.5. CHAPTER EXERCISES 120

recorded. The sample mean of the 10 observations was found to be 34.57 degrees
Celsius. The output for this data is:

One Sample t-test

data: oil
t = 554.0799, df = 9, p-value < 2.2e-16
alternative hypothesis: true mean is not equal to 0
95 percent confidence interval:
34.43033 34.71262
sample estimates:
mean of x
34.57147

Give an appropriate interpretation of the confdence interval found in the output,

in the context of the problem at hand.

40. Cadmium is a heavy metal that can build up to high levels in mushrooms. Re-
searchers investigated cadmium concentration in a specifc type of wild mushroom.
A random sample of 10 mushrooms was obtained, and it was found that the sam-
ple mean and sample standard deviation were 3.2 and 0.6, respectively. The units
are mg/kg of dry weight. Suppose it is reasonable to assume that the cadmium
concentrations are normally distributed.

(a) Calculate a 95% confdence interval for the population mean cadmium con-
centration for this type of mushroom.
(b) Give an interpretation of the interval in the context of the problem.
(c) Is it reasonable to simply assume the concentrations are normally distributed?
Should this assumption be investigated? If so, how? What are the conse-
quences if the concentrations are not normally distributed?

41. Researchers were interested in estimating the mean calcium level in the blood of
women in rural Guatemala. A sample of 18 women in a certain area of rural
Guatemala revealed a sample mean of 9.2 mg/dL, with a sample standard deviation
of 0.38 mg/dL. Assume for the purposes of these questions that the calcium levels
are approximately normally distributed, and the sample can be thought of as a
simple random sample of women from the area.

(a) What is a 95% confdence interval for the population mean?

(b) The correct interval 95% interval is (9.01, 9.39). Which one of the following is
the most appropriate interpretation of this interval?
i. We can be 95% confdent that the population mean calcium level of women
in this area lies between 9.01 and 9.39 mg/dL.
ii. We can be 95% confdent that the sample mean calcium level of the 18
women in the sample lies between 9.01 and 9.39 mg/dL.
iii. In repeated sampling, 95% of the sample means would lie between 9.01
Balka ISE 1.09 8.5. CHAPTER EXERCISES 121

and 9.39 mg/dL.

iv. We are more than 95% confdent that the population mean lies within the
interval, since 9.2 lies within the interval.
v. Since confdence intervals are two-sided, we can only be 90% confdent
that the population mean calcium level of women in this area lies between
9.01 and 9.39 mg/dL.

42. A researcher drew a random sample from a normally distributed population and
found a sample mean of 61.1. For this particular population, the population stan-
dard deviation was known to the researcher but the population mean was unknown.
The researcher found a 95% confdence interval for the population mean of (60.4,
61.8).

(a) What is the point estimate of µ?

(b) What is the margin of error?
(c) Calculate a 93% confdence interval for the population mean.
Balka ISE 1.09 8.5. CHAPTER EXERCISES 122
Chapter 9

Hypothesis Testing

J.B.’s strongly suggested exercises: 1, 5, 6, 10, 12, 13, 15, 19, 21, 33, 37, 38, 41, 48, 59,
61, 62, 63, 64, 65, 66

9.1 Introduction

1. Give appropriate hypotheses in each of the following scenarios.

(a) You wish to investigate whether there is a di˙erence in mean resting heart
rates of male and female Olympic athletes.
(b) You wish to investigate whether there are di˙erences in the six-month survival
rates of four di˙erent cancer treatments. (The six-month survival rate is the
true proportion of patients that survive for at least six months.)
(c) A producer of a type of protein bar claims their bars contain no more than
4.0 grams of saturated fat on average. You wish to show that the true mean
amount of saturated fat exceeds 4.0 grams.
(d) An auto manufacturer claims that at least 98% of purchasers of one of their
top models are satisfed with their purchase. You wish to show that the true
percentage is less than 98%.
(e) You wish to investigate whether there are di˙erences in the variances of the
gestation periods of three species of mammal.

123
Balka ISE 1.09 9.2. THE LOGIC OF HYPOTHESIS TESTING 124

9.2 The Logic of Hypothesis Testing

9.3 Hypothesis Tests for µ when σ is Known

9.3.1 Constructing Appropriate Hypotheses

2. Would it make sense to test the hypothesis H0 : X

¯ = 2? Why or why not?

9.3.2 The Test Statistic

3. Suppose we are interested in testing a null hypothesis about a population mean

µ. This population is normally distributed with a standard deviation of σ = 5.
Calculate the value of the appropriate test statistic in the following situations.

(a) n = 10, ¯
X = 15.5, µ0 = 20.
(b) n = 40, ¯
X = 15.5, µ0 = 20.
(c) n = 20, ¯
X = 35.5, µ0 = 30.
(d) n = 20, ¯
X = 65.5, µ0 = 30.

4. A researcher draws a random sample of 75 observations from a normally distributed

population. The mean of the population is unknown, but the standard deviation of
the population is known to be 6. The researcher wants to test the null hypothesis
that the mean of the population is 12, against the alternative that it is di˙erent
from 12. The researcher intends to use the test statistic:
¯ − 12
X
√
6/ 75

What is the sampling distribution of this test statistic if the null hypothesis is true?

9.3.3 The Rejection Region Approach to Hypothesis Testing

5. Suppose we wish to carry out a hypothesis test of H0 : µ = µ0 for a normally

distributed population where σ is known. In the following scenarios, what values
of the Z test statistic lead to a rejection of the null hypothesis? (What is the
appropriate rejection region?)

(a) Ha : µ =
6 µ0 , α = 0.10.
(b) Ha : µ < µ0 , α = 0.10.
Balka ISE 1.09 9.3. HYPOTHESIS TESTS FOR µ WHEN σ IS KNOWN 125

(c) Ha : µ > µ0 , α = 0.10.

(d) Ha : µ=6 µ0 , α = 0.004.
(e) Ha : µ < µ0 , α = 0.004.
(f) Ha : µ > µ0 , α = 0.004.

6. Test your conceptual understanding: Which of the following statements are true?
(You should be able to explain why a statement is true or why a statement is false.)

(a) The signifcance level of a test is the probability of rejecting the null hypothesis,
given the null hypothesis is true.
(b) If we reject the null hypothesis, then we know that the null hypothesis is false.
(c) If we reject the null hypothesis, then we know that the alternative hypothesis
is false.
(d) If we do not reject the null hypothesis, then we know that the null hypothesis
is true.
(e) If we do not reject the null hypothesis, then we know that the alternative
hypothesis is true.

9.3.4 P-values

7. Suppose we wish to test a null hypothesis about the mean of a normally distributed
population. Find the p-value in the following scenarios.

(a) Z = 2.76, Ha : µ =
6 10.
(b) Z = 2.76, Ha : µ < 10.
(c) Z = 2.76, Ha : µ > 10.
(d) Z = −1.74, Ha : µ =6 20.
(e) Z = −1.74, Ha : µ < 20.
(f) Z = −1.74, Ha : µ > 20.

8. Would the null hypothesis be rejected in the following scenarios?

(a) Ha : µ=6 10, p-value = 0.04, α = 0.05.

(b) Ha : µ > 10, p-value = 0.04, α = 0.05.
(c) Ha : µ= 6 10, p-value = 0.12, α = 0.10.
(d) Ha : µ > 10, p-value = 0.12, α = 0.10.
(e) Ha : µ < 10, p-value = 0.08, α = 0.10.

9. Suppose that for a certain population, the standard deviation is known but the
mean is unknown. A researcher draws a random sample of 600 observations from
this population, and fnds a sample mean of 2200. They carry out a Z test of the
null hypothesis that the population mean is 1000 against the alternative that it is
Balka ISE 1.09 9.3. HYPOTHESIS TESTS FOR µ WHEN σ IS KNOWN 126

greater than 1000, and fnd that the p-value is 0.08. Of the following options, which
one best describes the meaning of this p-value?

(a) If the null hypothesis is true, the probability of obtaining a sample mean at
least as large as 2200 is 0.08.
(b) If the null hypothesis is false, the probability of obtaining a sample mean at
least as large as 2200 is 0.08.
(c) The probability that the population mean is greater than 1000 is 0.08.
(d) The probability that the null hypothesis is false is 0.08.
(e) The probability that the null hypothesis is true is 0.08.

10. Test your conceptual understanding: Which of the following statements about p-
values are true? (You should be able to explain why a statement is true or why a
statement is false.)

(a) A p-value is a probability.

(b) We reject the null hypothesis when the p-value is less than or equal to the
signifcance level.
(c) The p-value of a test is the probability, given H0 is true, of obtaining the
observed value of the test statistic or a value with even greater evidence against
H0 and in favour of Ha .
(d) The p-value of a test depends on sample data and on the null and alternative
hypotheses.
(e) The p-value of a two-sided test will be greater than 1 if the null hypothesis is
true.

11. Test your conceptual understanding: Which of the following statements about p-
values are true? (You should be able to explain why a statement is true or why a
statement is false.)

(a) The p-value of a test will be less than 0.05 if the null hypothesis is false.
(b) The p-value of a test is the probability that the null hypothesis is true.
(c) If the p-value of a hypothesis test with a two-sided alternative is equal to
exactly 1, then we can be certain the null hypothesis is true.
(d) If the p-value of a hypothesis test with a two-sided alternative is equal to
exactly 1, then we can be certain the null hypothesis is false.
(e) If the p-value of a test is less than 0 that means there is extremely strong
evidence against the null hypothesis.
Balka ISE 1.09 9.4. EXAMPLES 127

9.4 Examples

12. A Slovenian study1 investigated the concentrations of cadmium and lead in edible
mushrooms in an area near the largest Slovenian thermal power plant. In one aspect
of the study, a sample of caps (n = 29) from the mushroom Boletus edulis revealed
a mean cadmium concentration of 9.26 mg/kg of dry weight. Suppose that for
samples of this type, σ = 5.00 mg/kg dw (this is an estimate based on this study,
but for the purposes of this question assume that the value of σ is known to be
5.00). Slovenian regulations give a tolerable value of 3 mg/kg dw (a cadmium level
of 3 mg/kg of dry weight or less is considered acceptable for consumption). Does
this sample provide strong evidence that the concentration of cadmium in this type
of mushroom in this area exceeds the tolerable value? Suppose we wish to carry out
a Z test of the appropriate hypothesis at the α = 0.05 signifcance level. (There
was some indication of a little right-skewness in the data, but the sample size here
is not very small, so it’s not unreasonable to use the Z test.)

(a) Give the appropriate hypotheses in words and symbols.

(b) What is the value of the appropriate test statistic?
(c) What is the p-value of the test?
(d) Is the null hypothesis rejected at the α = 0.05 level of signifcance?
(e) Give an appropriate conclusion to the hypothesis test.

13. A 2015 National Vital Statistics Report showed that in 2013, the weights of all
full-term newborn babies in the United States had a mean of 3.35 kg. Does the
mean weight of Hispanic newborns in the U.S. di˙er from this? A random sample
of 100 Hispanic newborns in 2013 had a mean of 3.318 kg. Suppose the standard
deviation of birth weights in the Hispanic population of the United States is known
to be σ = 0.47 kg. (This standard deviation is based on sample data, but it is
very close to the true value. For the purposes of this question, assume this is the
correct value of the population standard deviation.) Test the null hypothesis that
the mean weight of full-term Hispanic newborns in the U.S. in 2013 is the same as
that of the general population (3.35 kg).

(a) Give the appropriate hypotheses in words and symbols.

(b) What is the value of the appropriate test statistic?
(c) What is the p-value of the test?
(d) Is the null hypothesis rejected at the α = 0.05 level of signifcance?
(e) Give an appropriate conclusion to the hypothesis test.
1
Petkovšek, S. and Pokorny, B. (2013). Lead and cadmium in mushrooms from the vicinity of two large
emission sources in Slovenia. Science of the Total Environment, 443:944–954
Balka ISE 1.09 9.5. INTERPRETING THE P -VALUE 128

9.5 Interpreting the p-value

9.5.1 The Distribution of the p-value When H0 is True

14. Suppose we are about to draw a random sample from a normally distributed pop-
ulation and carry out a Z test of H0 : µ = 10.

(a) If the null hypothesis is true, what is the distribution of the p-value?
(b) If the null hypothesis is true, what is the average value of the p-value?

9.5.2 The Distribution of the p-value When H0 is False

15. Give an interpretation of the following p-values in terms of the evidence they give
against the null hypothesis. Suppose that there is no given value of α.

(a) 2.9 × 10−48

(b) 0.0008
(c) 0.02
(d) 0.07
(e) 0.12
(f) 0.45
(g) 0.88

9.6 Type I Errors, Type II Errors, and the Power of a

Test

16. Suppose we are testing H0 : µ = 10 against a two-sided alternative hypothesis at

α = 0.05.

(a) Suppose µ = 5, and when we carry out the test we fnd a p-value of 0.03. Will
we make a Type I error, a Type II error, or neither error?
(b) Suppose µ = 5, and when we carry out the test we fnd a p-value of 0.16. Will
we make a Type I error, a Type II error, or neither error?
(c) Suppose µ = 10, and when we carry out the test we fnd a p-value of 0.02.
Will we make a Type I error, a Type II error, or neither error?
(d) Suppose µ = 10, and when we carry out the test we fnd a p-value of 0.28.
Will we make a Type I error, a Type II error, or neither error?
Balka ISE 1.09 9.6. TYPE I ERRORS, TYPE II ERRORS, AND THE POWER OF A TEST 129

17. In a hypothesis test, under what conditions should a small value of α (0.00001, say)
be chosen? Under what conditions would it be more reasonable for α to be a larger
value (0.20, say)?

9.6.1 Calculating Power and the Probability of a Type II Error

18. Suppose we are about to sample 100 observations from a normally distributed pop-
ulation where it is known that σ = 20, but µ is unknown. We intend to test
H0 : µ = 30 against Ha : µ < 30 at α = 0.05.

(a) What values of the sample mean would lead to a rejection of the null hypoth-
esis?
(b) What is the power of the test if µ = 28?
(c) What is the power of the test if µ = 26?

19. Suppose we are about to sample 400 observations from a normally distributed pop-
ulation where it is known that σ = 160, but µ is unknown. We intend to test
H0 : µ = 800 against Ha : µ > 800 at α = 0.01.

(a) What values of the sample mean would lead to a rejection of the null hypoth-
esis?
(b) What is the power of the test if µ = 810?
(c) What is the power of the test if µ = 840?

20. (Warning: This question involves a power calculation for a test with a two-sided
alternative, and it is meaningful only if we do not wish to reach a directional
conclusion in the hypothesis test (we intend to either reject H0 or not reject H0 ).
Interpreting power can be tricky when the alternative hypothesis is two-sided and
we want to reach a directional conclusion.)

Suppose we are about to sample 50 observations from a normally distributed pop-

ulation where it is known that σ = 12, but µ is unknown. We intend to test
H0 : µ = 15 against Ha : µ =
6 15 at α = 0.05.

(a) What values of the sample mean would lead to a rejection of the null hypoth-
esis?
(b) What is the power of the test if µ = 16?
(c) What is the power of the test if µ = 17?
(d) What is the power of the test if µ = 13?

21. A researcher wants to test the null hypothesis that the mean of a certain population
is 0, against the alternative hypothesis that it is greater than 0. He has taken a
statistics course, but he forgets most of the material. He decides to take a random
Balka ISE 1.09 9.7. ONE-SIDED TEST OR TWO-SIDED TEST? 130

sample of size 50, and if the sample mean is larger than 10 he will reject the null
hypothesis. Suppose that the population is normally distributed and σ = 100.

(a) If µ = 0 what is the probability the researcher commits a Type I error?

(b) If µ = 0 what is the probability the researcher commits a Type II error?
(c) If µ = 5 what is the probability the researcher commits a Type I error?
(d) What is of the power of the test if µ = 5?
(e) What is of the power of the test if µ = 10?

9.6.2 What Factors A˙ect the Power of the Test?

22. All other factors being equal, what is the e˙ect of the following on the power of the
test?

(a) An increase in the sample size.

(b) A decrease in the signifcance level.
(c) An increase in the distance between the true mean and the hypothesized mean.
(d) A decease in the population variance.

9.7 One-Sided Test or Two-Sided Test?

23. The mean of a certain population is thought to be 1000. We wish to draw a sample
and test the null hypothesis that the mean is 1000. In two or three sentences, state
the advantages and disadvantages of choosing a one-sided alternative hypothesis
instead of a two-sided alternative.

24. Suppose it is commonly believed that the mean of a certain population is 40. You
feel that the mean is actually much larger than 40, so you draw a sample and carry
out a one-sided hypothesis test at the 5% level of signifcance, using the alternative
hypothesis Ha : µ > 40. You fnd that the sample mean is actually much smaller
than 40, and you obtain a p-value of 0.99. You now decide to simply change your
mind, abandon your earlier hypothesis and, using the same data, test the alternative
hypothesis Ha : µ < 40. You fnd and report a p-value of 0.01, and reject the null
hypothesis at the 5% signifcance level. Is this acceptable statistical practice?
Balka ISE 1.09 9.8. STATISTICAL SIGNIFICANCE AND PRACTICAL SIGNIFICANCE 131

9.8 Statistical Signifcance and Practical Signifcance

25. Label the following statements as true or false. (You should be able to explain why
a statement is true or why a statement is false.)

(a) If the p-value of a hypothesis test is 0.0000001, then there is very strong
evidence that the results of the test have practical importance.
(b) If the sample size is extremely large, then even tiny, meaningless di˙erences
from the hypothesized value will be found statistically signifcant.
(c) If the results of a hypothesis test are reported, then there is no need to also
report the associated confdence interval for the parameter.

9.9 The Relationship Between Hypothesis Tests and Con-

fdence Intervals

26. The p-value for a two-sided hypothesis test of H0 : µ = 50 is found to be 0.0233.

(a) Would a 95% confdence interval for µ contain 50?

(b) Would a 99% confdence interval for µ contain 50?

27. Suppose a 95% confdence interval for a population mean is found to be (1.17, 8.23).
For the same data, what can be said of the p-value of a two-sided test of the null
hypothesis that the population mean is 0?

(a) The p-value is greater than 0.05.

(b) The p-value is less than 0.05.
(c) The p-value is equal to 0.05.
(d) It is impossible to determine the p-value with the given information.

9.10 Hypothesis Tests for µ When σ is Unknown

28. Suppose we wish to carry out a hypothesis test of H0 : µ = µ0 for a normally

distributed population where σ is unknown. In the following scenarios, what values
of the t test statistic will lead to a rejection of the null hypothesis? (What is the
appropriate rejection region?)

(a) Ha : µ =
6 µ0 , α = 0.10, n = 15.
(b) Ha : µ < µ0 , α = 0.10, n = 15.
(c) Ha : µ > µ0 , α = 0.10, n = 15.
Balka ISE 1.09 9.11. MORE ON ASSUMPTIONS 132

(d) Ha : µ =
6 µ0 , α = 0.05, n = 27.
(e) Ha : µ < µ0 , α = 0.001, n = 91.
(f) Ha : µ > µ0 , α = 0.15, n = 348.

29. Suppose we are testing a hypothesis about the mean of a normally distributed
population. Find the p-value in each of the following situations.

(a) t = 2.76, n = 5, Ha : µ =
6 10.
(b) t = 2.76, n = 5, Ha : µ < 10.
(c) t = 2.76, n = 5, Ha : µ > 10.
(d) t = −1.84, n = 18, Ha : µ =
6 20.
(e) t = −1.84, n = 18, Ha : µ < 20.
(f) t = −1.84, n = 18, Ha : µ > 20.

9.10.1 Examples of Hypothesis Tests Using the t Statistic

30. The nutrition information published by a major fast-food chain claims that their
bean burritos contain 11 grams of fat on average. A sample of 6 of these burritos
were analyzed, and it was found that they contained an average of 11.49 g of fat
with a standard deviation of 1.45 g. (This information is based on sample data
published in the US National Nutrient Database.) Suppose we wish to test the
null hypothesis that the published nutrition information is correct, and that these
burritos contain 11.0 g of fat on average. (The use of the t procedures is always a
little questionable for such a small sample size (n = 6), but there were no outliers
in the data so the use the t procedure is not completely out of line.)

(a) In words and symbols, what are the hypotheses of the appropriate hypothesis
test?
(b) What is the value of the appropriate test statistic?
(c) What is the p-value of the test?
(d) Give an appropriate conclusion to the hypothesis test.
(e) What assumptions are necessary in order for the test to be valid? What biases
might be present?

9.11 More on Assumptions

31. A sample of 20 values resulted in the normal quantile plot seen in Figure 9.1. Would
it be reasonable to use the t procedures on this sample data?
Balka ISE 1.09 9.12. CHAPTER SUMMARY 133

12
10
8
sa m p le d va lu e s

6
4
2
0
-2 -1 0 1 2
Q u a n ti l e s o f S ta n d a rd N o rm a l

Figure 9.1: Normal QQ plot for Question 31.

32. Suppose we are about to sample from a population and carry out a two-sided t test
about a population mean µ.

(a) For what shapes of distribution does the t test perform poorly?
(b) If we use the t test when we are sampling from a strongly skewed distribution,
what are the implications?

9.12 Chapter Summary

9.12.1 Basic Calculations

33. Suppose we are interested in testing a hypothesis about the mean of a normally
distributed population. Calculate the value of the appropriate test statistic in
the following scenarios. Also, assuming that the null hypothesis is true, state the
distribution of the test statistic in each scenario.

(a) n = 10, ¯
X = 15.5, s = 2, µ0 = 20.
(b) n = 10, ¯
X = 15.5, s = 20, µ0 = 20.
(c) n = 40, ¯
X = 35.5, σ = 2, µ0 = 30.
(d) n = 50, ¯
X = 45.5, s = 20, µ0 = 40.

34. Suppose we are interested in testing a hypothesis about the mean of a normally
distributed population. Calculate the value of the appropriate test statistic in
the following scenarios. Also, assuming that the null hypothesis is true, state the
distribution of the test statistic in each scenario.

(a) n = 20, X
¯ = 35.5, s = 2, µ0 = 30.
(b) n = 25, X
¯ = 35.5, σ = 20, µ0 = 40.
Balka ISE 1.09 9.12. CHAPTER SUMMARY 134

(c) n = 200, X¯ = 65.5, s = 2, µ0 = 30.

(d) n = 60, 000, X
¯ = 32.1, s = 10, µ0 = 32.

35. Consider the following output for a one-sample t procedure.

One-sample t-Test
data: madeupdata
t = -4.1027, df = 24, p-value = 0.0004
alternative hypothesis: mean is not equal to 50
95 percent confidence interval:
46.80365 48.94322
sample estimates:
mean of x
47.87344

Which of the following statements are true?

(a) The null hypothesis is X¯ = 50.

(b) The sample size is 24.
(c) The area to the left of −4.1027 under a t distribution with 24 degrees of
freedom is 2 × 0.0004 = 0.0008.
(d) If we used this data to test the null hypothesis that the population mean is
40.0, against a two-sided alternative hypothesis, the p-value would be less than
0.05.
(e) The 95% margin of error is greater than 20.0.

9.12.2 Concepts

36. What are the assumptions of a Z test on the population mean µ?

37. What are the assumptions of a t test on the population mean µ?

38. If the assumptions of a Z test or t test are violated, what are the consequences?

39. In inference procedures for a population mean µ, when is a t test more appropriate
than a z test?

40. Which of the following should be based on the current sample’s data?

(a) The signifcance level of a test.

(b) The hypothesized mean.
(c) The choice of alternative hypothesis.
(d) The sample mean.
Balka ISE 1.09 9.12. CHAPTER SUMMARY 135

(e) The value of the test statistic.

(f) The p-value.

41. Suppose we are about to sample from a normally distributed population and con-
duct a one-sample t test of H0 : µ = 250.

(a) If the null hypothesis is true, what is the average value of the test statistic?
(b) If the null hypothesis is true, what is the average value of the p-value?

42. Suppose we are about to carry out 25 independent hypothesis tests. We intend to
carry out each test at a signifcance level of α = 0.05. Suppose that the assumptions
of each test are true, and, unknown to us, the null hypothesis is true in all 25
situations.

(a) What is the distribution of the number of Type I errors that we will make in
these 25 tests?
(b) On average, how many Type I errors will we make?
(c) What is the probability that we make at least one Type I error?

43. A researcher wants to test the null hypothesis H0 : µ = 29 against the alternative
that µ is di˙erent from 29. She forgets how to carry out the test, but remembers
how to calculate confdence intervals. She fnds a 95% confdence interval for µ of
(23.20, 28.80), and a 99% confdence interval for µ of (22.32, 29.68). Which one of
the following could be the p-value of the appropriate test?

(a) 0.0022.
(b) 0.0358.
(c) 0.0517.
(d) 0.0812.
(e) 1.8143.

44. Consider the following output from a statistical computing package.

One-sample t-Test
data: mydata
t = -0.2518, df = 39, p-value = 0.8025
alternative hypothesis: true mean is not equal to 0
95 percent confidence interval:
-0.3412277 0.2656748

(a) What is the value of the sample mean?

(b) What is the value of the margin of error?
(c) What would the p-value be if the alternative was µ < 0 instead of µ 6= 0?
(Don’t look up a range of values in a table—base your response on the output.)
Balka ISE 1.09 9.12. CHAPTER SUMMARY 136

(d) What would the p-value be if the alternative was µ > 0 instead of µ 6= 0?
(e) Suppose the p-value in the original output was 0.008025 instead of 0.8025.
Give an appropriate conclusion to the t test.

45. A researcher is interested in testing the null hypothesis that the mean of a certain
population is 32. She strongly believes that the population mean is less than 32,
and she feels that a one-sided alternative hypothesis is appropriate. She draws
a simple random sample of size 15, and uses software to perform a t-test with a
one-sided alternative hypothesis. The following output summarizes the results.

One-sample t-Test
data: madeupdata
t = -2.6839, df = 14, p-value = 0.0089
alternative hypothesis: mean is less than 32

Consider the 3 values: 28.0, 29.0, 34.0. Which of these values could possibly be the
true value of µ?

46. Consider the two normal quantile-quantile plots illustrated in Figure 9.2.
4

1.5
1.0
Sample Quantiles

Sample Quantiles
3

0.5
2

0.0
-0.5
1

-1.0
-1.5
0

-2 -1 0 1 2 -1.0 -0.5 0.0 0.5 1.0

Theoretical Quantiles Theoretical Quantiles

Figure 9.2

The plot on the left is for a sample of size 100. The plot on the right is for a sample
of size 5. In each of these situations, would it be reasonable to use the t procedures
to conduct tests and construct confdence intervals?

47. Consider the following output for a one sample problem.

Balka ISE 1.09 9.12. CHAPTER SUMMARY 137

One Sample t-test

data: mydata
t = 4.1611, df = 39, p-value = 0.0001687
alternative hypothesis: true mean is not equal to 50
95 percent confidence interval:
51.26444 53.65645
sample estimates:
mean of x
52.46044

(a) The sample standard deviation for this data is 3.739676. What is the value of
the standard error of the sample mean?
(b) Give an appropriate conclusion to the hypothesis test.
(c) Suppose that before they gathered the data, the researchers in this problem
decided that a 5% signifcance level was appropriate in this situation. Suppose
also that the population is normally distributed and the null hypothesis is
true. Before the data was gathered, what was the probability that this test
would result in a Type I error?
(d) Suppose that the true value of the population mean is actually 50. If we make
an appropriate conclusion to the hypothesis test at the 5% signifcance level
(based on the output), would we make a Type I error, a Type II error, or
neither error?
(e) The data for this problem is plotted in Figure 9.3. Do these plots give any
indication that the assumptions of the t procedure are violated?

60

60
55
55
Sample Quantiles

50
50

45 45
-2 -1 0 1 2
Theoretical Quantiles

Figure 9.3: Plots for Question 47e.

48. Figure 9.4 illustrates boxplots for 3 samples (labelled A, B, and C). Each sample
has a sample size of 50.

Suppose that for each sample, we carry out a hypothesis test of H0 : µ = 18 against
Balka ISE 1.09 9.12. CHAPTER SUMMARY 138

30

25

20

15

10
A B C

Figure 9.4: Boxplots for three samples.

a two-sided alternative hypothesis.

(a) Which one of the three samples would have the largest p-value of the test?
(b) Which one of the three samples would have the smallest p-value of the test?

49. Suppose that we run two experiments, and then carry out a hypothesis test for
the data resulting from each experiment. For the data of the frst experiment, the
p-value is found to be 0.100000001, and for the data of the second experiment, the
p-value is found to be 0.09999999999.

(a) Is the null hypothesis rejected in Experiment 1 at α = 0.10?

(b) Is the null hypothesis rejected in Experiment 2 at α = 0.10?
(c) Which one of the following statements is the best assessment of the strength
of the evidence against the null hypothesis in the two experiments?
i. Experiment 1 shows stronger evidence against the null hypothesis.
ii. Experiment 2 shows stronger evidence against the null hypothesis.
iii. The evidence against the null hypothesis is essentially the same for both
experiments.

50. In hypothesis testing, under certain conditions we should choose α to be a very

small value. Under which of the following situations should we choose a very small
value for α?

(a) If a Type I error would cost thousands of lives, but a Type II error would
result in a loss of $2.
(b) If a Type II error would cost thousands of lives, but a Type I error would
result in a loss of $2.
(c) If we want to be very sure the null hypothesis is true.
(d) If we want the test to have very high power.
Balka ISE 1.09 9.12. CHAPTER SUMMARY 139

51. Suppose that you fnd a coin on the street that is strange looking and is obviously
not regular currency. This coin may very well be a trick coin that comes up heads
more (or less) often than tails. Suppose that you want to test the null hypothesis
that this coin is a fair coin (the probability it comes up heads when tossed is 0.5),
against the alternative that it comes up tails more often than heads on average.
You foolishly decide to simply toss the coin once, and if it comes up tails, you will
decide that it is biased toward tails.

(a) If the null hypothesis is true, what is the probability of a Type I error?
(b) Suppose that the null hypothesis is false. For what true probability of tails
would it be impossible to make a Type II error?

52. The following analysis of a hypothesis test contains at least one glaring error.

A researcher wants to test the null hypothesis H0 : X ¯ = 10, against the alternative
that it is less than 10. The researcher performs a t test, and observes a test statistic
of −3.84. The p-value was found to be 1.12, so the null hypothesis was not rejected
at the 5% level of signifcance.

What errors are in this analysis?

53. A researcher draws a large sample (several hundred observations) from a normally
distributed population, where it is known that σ = 10. She wants to test the null
hypothesis that µ = 50, against the alternative that it is less than 50. Unknown to
the researcher, the true value of µ is 37. The researcher carries out a Z test, and
fnds that the p-value of the test is 0.5. What was the value of the sample mean?
(If it is impossible to determine, say so.)

54. A pet food manufacturer suspects that one of its machines is overflling bags of
dog food that are supposed to contain 20 kg of food. They want to test the null
hypothesis that the mean fll is 20 kg against the alternative hypothesis that it
is greater than 20 kg. They don’t know much about statistics, and decide that
they will reject the null hypothesis if the sample mean of 100 randomly selected
bags is greater than 20.1 kg. Unknown to the manufacturer, the machine is truly
overflling the bags, with a mean of 20.3 kg. When the manufacturer conducts the
test and draws a conclusion, what is the probability that they make a Type II error?
(Assume that the population standard deviation is known to be 0.5 kg.)

55. A scientist is about to carry out an experiment and then conduct a Z test on her
data. She vaguely recalls how to calculate the appropriate Z test statistic, and
remembers that the vast majority of randomly selected Z values fall between −3
and +3. She decides that she will reject the null hypothesis if the value of the Z
test statistic falls outside of this range.

(a) If the null hypothesis is true, what is the probability that she makes a Type I
Balka ISE 1.09 9.12. CHAPTER SUMMARY 140

error?
(b) If the null hypothesis is true (and she keeps the same rejection rule), what will
happen to the probability of a Type I error as the sample size increases?
(c) If the null hypothesis is false, what is the probability that she makes a Type
I error?
(d) If the null hypothesis is false (and she keeps the same rejection rule), what
will happen to the power of the test as the sample size increases?

56. Suppose we should be using a t test statistic to test a null hypothesis, against a
two-sided alternative. We properly calculate the t statistic, but we make a mistake
fnd the p-value of the test from the standard normal distribution. What e˙ect will
that have on the reported p-value?

(a) The reported p-value will be greater than it should be.

(b) The reported p-value will be less than it should be.
(c) The reported p-value will be equal to what it should be.
(d) The e˙ect on the p-value is impossible to determine.

57. Test your conceptual understanding: Which of the following statements are true?
(You should be able to explain why a statement is true or why a statement is false.)

(a) All else being equal, the power of a test will increase as the population variance
decreases.
(b) If the signifcance level of a test is decreased (from 0.10 to 0.05, say), then the
probability of a Type II error will increase.
(c) The probability of a Type II error decreases the farther the true population
mean is from the hypothesized mean.
(d) If the null hypothesis is true, then we cannot make a Type II error.

58. Test your conceptual understanding: Which of the following statements are true?
(You should be able to explain why a statement is true or why a statement is false.)

(a) Power = 1 − P (Type I error).

(b) Power = 1 − P (Type II error).
(c) P (Type I error) = 1 − P (Type II error).
(d) All else being equal, the power of a test will decrease the closer the true value
of the population mean is to the hypothesized value of the population mean.
(e) All else being equal, the power of a test will increase as the sample size in-
creases.

59. Test your conceptual understanding: Which of the following statements are true?
(You should be able to explain why a statement is true or why a statement is false.)

(a) If a null hypothesis is rejected at the 5% signifcance level, then it would also
Balka ISE 1.09 9.12. CHAPTER SUMMARY 141

be rejected at the 10% signifcance level.

(b) If a null hypothesis is rejected at α = 0.05, then it would also be rejected at
α = 0.01.
(c) If a null hypothesis is not rejected at the 5% signifcance level, then it will
defnitely not be rejected at the 10% signifcance level.
(d) The signifcance level of a test is 1 − p-value.
(e) The signifcance level of a test is the probability that the null hypothesis is
true.

60. Test your conceptual understanding: Which of the following statements about p-
values are true? (You should be able to explain why a statement is true or why a
statement is false.)

(a) If the p-value of a hypothesis test with a two-sided alternative is equal to

exactly 0, we can be certain the null hypothesis is false. (This is a conceptual
question—for all the tests we have done, it is not possible to get a p-value of
exactly 0.)
(b) The p-value of a test is the probability of getting an insignifcant result, given
the null hypothesis is true.
(c) The p-value of a test is less than the signifcance level whenever the null hy-
pothesis is false.
(d) The p-value of a test is the probability of rejecting the null hypothesis when
it is false.
(e) If the p-value of a test is large (0.99, say), then we can be almost positive that
the null hypothesis is true.

61. Test your conceptual understanding: Which of the following statements about p-
values are true? (You should be able to explain why a statement is true or why a
statement is false.)

(a) For a two-sided alternative hypothesis, the p-value of the test will equal 1 if
the sample mean equals the hypothesized mean.
(b) If we are sampling from a normally distributed population and conducting
an ordinary t test, then if the null hypothesis is true the p-value will have a
uniform distribution between 0 and 1.
(c) If the null hypothesis is true, then on average the p-value will equal 0.5.
(d) If the null hypothesis is false, then on average the p-value will equal 0.5.
Balka ISE 1.09 9.12. CHAPTER SUMMARY 142

9.12.3 Applications

Weight
62. Body Mass Index ( Height2 , where weight is in kg and height is in m) is a measure

of body shape. A study2 investigated BMI in Texas youth. Student participants

were recruited from 6 schools in a school district in a rural area of North Central
Texas (about 100 km from Fort Worth). Students participated only if their parents
provided informed consent. Students who participated were given a free T-shirt.

Here will look only at the BMI data for the 94 14 year-old boys in the study. For
the boys in this sample, the mean BMI is 23.3, with a standard deviation of 5.4. A
study by the Centre for Disease Control3 at approximately the same time showed
that the average BMI for 14 year old boys in the United States was 22.4 kg/m2 .
(The CDC study involved a sample, but for the purposes of this question assume
that the true mean BMI for 14 year-old boys in the United States is 22.4.) Suppose
we wish to test whether the study by Duran et al. gives strong evidence that the
mean BMI for 14 year-old boys in in rural Texas di˙ers from the BMI for 14 year-old
boys in the United States.

Figure 9.5 illustrates the BMI values for the 94 boys. The red dashed line on the
boxplot indicates the hypothesized value of 22.4.
40

40
35

35
Sample Quantiles
30

30
BMI
25

25
20

20
15

15

-2 -1 0 1 2
Theoretical Quantiles

(a) The line represents the hypothesized value of (b) Normal quantile-quantile plot.
22.4.

Figure 9.5: BMI for 94 boys in rural Texas.

(a) Do these plots show a violation of the normality assumption? If there is a

violation, is it a serious problem in this situation? (Can we still use the t
procedures?)
2
Duran et al. (2013). Growth and weight status of rural texas school youth. American Journal of Human
Biology, 25:71–77. The data used here is simulated data with the same summary statistics as given in
their Table 1.
3
Fryar et al. (2012). Anthropometric reference data for children and adults: United states, 2007–2010.
National Center for Health Statistics. Vital Health Stat., 11(252)
Balka ISE 1.09 9.12. CHAPTER SUMMARY 143

(b) In words and symbols, what are the hypotheses of the appropriate hypothesis
test?
(c) What is the value of the appropriate test statistic?
(d) What is the p-value of the test?
(e) Give an appropriate conclusion to the hypothesis test.
(f) The output from the statistical software R for this data is:
data: BMI
t = 1.6159, df = 93, p-value = 0.1095
alternative hypothesis: true mean is not equal to 22.4
95 percent confidence interval:
22.19397 24.40603
sample estimates:
mean of x
23.3
Give a summary of the results of the analysis (including the results of the
hypothesis test and the confdence interval). To what population do your
conclusions apply? Comment on any biases that might be present.

63. Accurately estimating the age at death of unidentifed human remains is important
in forensic science. Is there bias in the procedures used to estimate age at death?
A study4 investigated the validity of 3 dental methods for estimating age at death.
In one part of the study, the Bang and Ramm (BR) method was used to estimate
the ages of 11 individuals that had a known age of 22. (These were individuals that
donated their bodies for science and medical education, or they were from remains
in the Weisbach Collection, which is a collection of skeletal material of known age
and sex.) The di˙erence between the BR estimate and the true age was recorded
(BR estimate − true age), and the average di˙erence was found to be 11.073 years
and the standard deviation was found to be 7.784 years. The results are illustrated
in Figure 9.6.

If the BR estimation method is unbiased in this situation (the estimated age is

correct, on average), then the true (theoretical) mean di˙erence is 0. Does this
sample give strong evidence of a bias? Let’s investigate this by carrying out a
test of the null hypothesis that the true mean di˙erence is 0 against a two-sided
alternative hypothesis.

(a) Do these plots show a violation of the normality assumption? Is the use of the
t procedures reasonable?
4
Meinl et al. (2008). Comparison of the validity of three dental methods for the estimation of age at
death. Forensic Science International, 178:96–105. Values in this question are estimated from their
Figure 2. Since the bias of the di˙erent aging techniques depends on the real age, this study looked at
teeth belonging to bodies of various ages at death. In this question we are looking only at individuals
with a true age of 22.
Balka ISE 1.09 9.12. CHAPTER SUMMARY 144

20

20
BR estimate - true age

Sample Quantiles
15

15
10

10
5

5
0

0
-1.5 -1.0 -0.5 0.0 0.5 1.0 1.5
Theoretical Quantiles

(a) The line represents the hypothesized value of (b) Normal quantile-quantile plot.
0.

Figure 9.6: The di˙erence between the estimated age and true age for 11 22 year-olds.

(b) In words and symbols, what are the hypotheses of the appropriate hypothesis
test?
(c) What is the value of the appropriate test statistic?
(d) What is the p-value of the test?
(e) Give an appropriate conclusion to the hypothesis test.
(f) The output from the statistical software R for this data is:
data: age_at_death
t = 4.718, df = 10, p-value = 0.0008189
alternative hypothesis: true mean is not equal to 0
95 percent confidence interval:
5.843638 16.302362
sample estimates:
mean of x
11.073
Give a summary of the results of the analysis (including the results of the
hypothesis test and the confdence interval). To what population do your
conclusions apply? Comment on any sampling biases that might be present.

64. The ratio of the length of the index fnger to the length of the ring fnger is called the
2D:4D ratio. A study5 investigated the 2D:4D ratio in students in an introductory
biology course at Western Washington University. The distribution of the 2D:4D
ratio is di˙erent for groups of di˙erent ethnic backgrounds, and this study restricted
the sample to people of European descent. In one part of the study, a sample of
135 female students had an average 2D:4D ratio on the left hand of 0.994 and a
standard deviation of 0.035.

A question of interest is whether the mean 2D:4D ratio di˙ers from 1. (If an
5
Stevenson et al. (2007). Attention defcit/hyperactivity disorder (ADHD) symptoms and digit ratios in
a college sample. American Journal of Human Biology, 19:41–50
Balka ISE 1.09 9.12. CHAPTER SUMMARY 145

individual has a 2D:4D ratio of 1, then their index fnger and ring fnger are of
equal length.) Test the null hypothesis that the population mean 2D:4D ratio
is 1, against the alternative hypothesis that it di˙ers from 1. (2D:4D ratios are
approximately normally distributed, and in addition the sample size is large here,
so the t procedure is appropriate.)

(a) In words and symbols, what are the hypotheses of the appropriate hypothesis
test?
(b) What is the value of the appropriate test statistic?
(c) What is the p-value of the test?
(d) Give an appropriate conclusion to the hypothesis test at α = 0.05.
(e) The output from the statistical software R for this data is:
data: 2D4D
t = -1.9918, df = 134, p-value = 0.04842
alternative hypothesis: true mean is not equal to 1
95 percent confidence interval:
0.9880422 0.9999578
sample estimates:
mean of x
0.994
Give a summary of the results of the analysis (including the results of the
hypothesis test and the confdence interval). To what population do your
conclusions apply? Comment on any biases that might be present.

65. The nutrition information published by a major fast-food chain claims that their
quarter-pound cheeseburgers contain 30 grams of protein. A sample of 6 of these
cheeseburgers was analyzed, and it was found that they contained an average of
33.81 g of protein with a standard deviation of 2.72 g. (This information is based
on sample data published in the U.S. National Nutrient Database.) Suppose we
wish to test the null hypothesis that the published nutrition information is correct,
and that these cheeseburgers contain 30 g of protein on average. (The use of the t
procedures is always a little questionable for such a small sample size (n = 6), but
there were no outliers in the data so the use the t procedure is not completely out
of line.)

(a) In words and symbols, what are the hypotheses of the appropriate hypothesis
test?
(b) What is the value of the appropriate test statistic?
(c) What is the p-value of the test?
(d) Give an appropriate conclusion to the hypothesis test at α = 0.05.
(e) What assumptions are necessary in order for the test to be valid? What biases
might be present?
Balka ISE 1.09 9.12. CHAPTER SUMMARY 146

66. A study6 investigated concentrations of cadmium and lead in edible mushrooms

in an area near the largest Slovenian thermal power plant. Slovenian regulations
give a tolerable lead value of 5 mg/kg dw (a lead level of 5 mg/kg of dry weight
or less is considered acceptable for consumption). Suppose that the researchers
suspected that the lead concentrations in mushrooms in this area might exceed this
level, and they wanted to see if they had strong evidence of this. In one aspect
of the study, a sample of caps (n = 21) from the mushroom Laccaria amethystina
(commonly known as the Amethyst Deceiver) revealed a mean lead concentration
of 0.94 mg/kg dw and a standard deviation of 0.25 mg/kg dw. Does this sample
provide strong evidence that the mean lead concentration in this type of mushroom
in this area exceeds the tolerable value? Suppose we wish to carry out a t test of
the appropriate hypothesis.

(a) Give the appropriate hypotheses in words and symbols.

(b) What is the value of the appropriate test statistic?
(c) What is the p-value of the test?
(d) Give an appropriate conclusion to the hypothesis test.

67. Franklin et al. (2012)7 investigated various aspects of tandem running in ants. Tan-
dem running is a form of recruitment in which one ant with knowledge of the loca-
tion of a food source or new nest site leads another ant to that location. (Optional:
Watch an example of tandem running here: (1:53) (http://www.youtube.com/watch?v=X2C7Sy2oPik))
Tandem running can be thought of as a form of teaching and learning.

The study investigated various factors associated with tandem running, and inves-
tigated whether the age of the ant and the experience of the ant had any e˙ect on
tandem running. Ants (Temnothorax albipennis) were categorized into 4 categories:
Young and inexperienced (YI), young and experienced (YE), old and inexperienced
(OI), old and experienced (OE). One aspect of the study investigated the speed of
tandem running for 51 tandem runs in which a YE ant was leading a YI ant. The
mean speed (mm/s) of the run was recorded. Figure 9.7 illustrates the speed of the
runs. For these 51 runs the sample mean was 1.80 mm/s, and the sample standard
deviation was 0.60 mm/s.

Suppose it was previously believed that under the conditions of this study, the true
mean speed of the tandem running would be 1.0 mm/s. Does this study yield strong
evidence against this claim?

(a) Do these plots show a violation of the normality assumption? If there is a

6
Petkovšek, S. and Pokorny, B. (2013). Lead and cadmium in mushrooms from the vicinity of two large
emission sources in Slovenia. Science of the Total Environment, 443:944–954
7
Franklin et al. (2012). Do ants need to be old and experienced to teach? The Journal of Experimental
Biology, 215:1287–1292 The data used here is simulated data based on their Figure 2, with similar results
and conclusions.
Balka ISE 1.09 9.12. CHAPTER SUMMARY 147

2.5

2.5
Sample Quantiles
Speed (mm/s)
2.0

2.0
1.5

1.5
1.0

1.0
0.5

0.5
-2 -1 0 1 2
Theoretical Quantiles

(a) The line represents the hypothesized value of (b) Normal quantile-quantile plot.
1.0.

Figure 9.7: Speed of 51 tandem runs.

violation, is it a serious problem in this situation? (Can we still use the t

procedures?)
(b) In words and symbols, what are the hypotheses of the appropriate hypothesis
test?
(c) What is the value of the appropriate test statistic?
(d) What is the p-value of the test?
(e) Give an appropriate conclusion to the hypothesis test.
(f) The output from the statistical software R for this data is:
data: Tandem running
t = 9.5219, df = 50, p-value = 8.14e-13
alternative hypothesis: true mean is not equal to 1
95 percent confidence interval:
1.631247 1.968753
sample estimates:
mean of x
1.8
Give a summary of the results of the analysis (including the results of the
hypothesis test and the confdence interval). To what population do your
conclusions apply? Comment on any biases that might be present.

9.12.4 Extra Practice Questions

68. A study by the Center for Science in the Public Interest found that movie theatre
popcorn is often very high in calories and saturated fat. The CSPI found that on
top of the poor stated nutritional characteristics, in reality the popcorn had much
higher calorie and fat content than what was claimed.

Suppose you wish to investigate the calorie content of large bags of popcorn at your
Balka ISE 1.09 9.12. CHAPTER SUMMARY 148

local theatre, where the theatre chain claims that a large untopped bag of their
popcorn contains 920 calories on average. You purchase 14 bags of this type of
popcorn, and suppose that it is reasonable to think that these 14 bags represent a
random sample from the population of bags of this type. You have these bags of
popcorn analyzed and fnd that the mean calorie content is 1082 and the standard
deviation is 60.

A normal quantile-quantile plot and boxplot of the calorie counts are given in
Figure 9.8. The theatre’s claimed value of 920 calories is represented by a line on
the boxplot.

1200

1150
1150

1100
1100
Sample Quantiles
Calories

1050 1050

1000

950
1000

900
-1 0 1

Theoretical Quantiles

(a) Boxplot of calorie count. The hypothe- (b) Normal QQ plot of calorie count.
sized value of 920 is indicated by the dotted
line.

Figure 9.8: Boxplot and Normal QQ plot of calorie counts.

(a) Do these plots give any indication that the t procedures should not be used
here?
(b) Construct a 90% confdence interval for the population mean calorie content
of bags of this type.
(c) Suppose we wish to carry out a test of the null hypothesis that the population
mean calorie content is what the company claims it to be. If we feel a one-
sided alternative hypothesis is most appropriate in this situation, what are the
appropriate hypotheses?
(d) Which procedure (z or t) is appropriate here? Why?
(e) What is the value of the appropriate test statistic?
(f) What is the p-value of the test?
(g) Give an appropriate summary of the results of the analysis.

69. The Environment Canada safety guideline for arsenic in soil is 12.0 ppm. (If the
Balka ISE 1.09 9.12. CHAPTER SUMMARY 149

level of arsenic in soil is no more than 12.0 parts per million, the soil is considered
safe.) Suppose we suspect that the average arsenic level in the soil in playgrounds of
elementary schools in a large school district is greater than 12.0 ppm. We randomly
sample 9 schools, take a soil sample from each school, and fnd that the average
arsenic level is 16.7 ppm with a standard deviation of 4.8 ppm. Does this sample
provide strong evidence that the true mean arsenic level is greater than 12.0 ppm?
(While arsenic levels in soil likely have a right-skewed distribution, for the purposes
of this question assume they are normally distributed.)

(a) What are the appropriate hypotheses of the test?

i. H0 : X
¯ = 12.0, Ha : X¯ 6= 12.0
ii. H0 : µ = 12.0, Ha : µ > 12.0
iii. H0 : µ = 12.0, Ha : µ =
6 12.0
iv. H0 : µ = 12.0, Ha : µ < 12.0
v. All of the above are appropriate.
(b) Calculate the appropriate test statistic and p-value.
(c) Which one of the following is the most appropriate conclusion at the α = 0.05
level of signifcance?
i. There is signifcant evidence that the population mean arsenic level is
within Environment Canada guidelines.
ii. There is signifcant evidence that the population mean arsenic level is
greater than the Environment Canada guideline.
iii. There is signifcant evidence that the sample mean arsenic level is within
the Environment Canada guideline.
iv. We know for certain that the population mean arsenic level is greater than
the Environment Canada guideline.
(d) Why might this not be the most appropriate test to consider in this situation?
(e) Suppose that the population mean arsenic level is in fact exactly equal to 12.0
ppm. In our hypothesis test, did we make a Type I error, a Type II error, or
neither error?

70. Researchers are investigating possible side e˙ects of a newly developed drug. As
part of their investigation, the researchers conduct an experiment to see if this drug
has an e˙ect on the response time of rats. The researchers know from a great deal
of past experience that the average response time for rats that are not given the
drug is 1.25 seconds. The researchers administer the drug to a sample of 40 rats and
measure the response time. They use the data to test the null hypothesis that the
mean response time is 1.25 against a two-sided alternative, and obtain the following
output.
Balka ISE 1.09 9.12. CHAPTER SUMMARY 150

One Sample t-test

data: rat
t = 8.8286, df = 39, p-value = 7.704e-11
alternative hypothesis: true mean is not equal to 1.25
95 percent confidence interval:
1.424516 1.528247
sample estimates:
mean of x
1.476382

(a) Give a summary of the analysis.

(b) Is it wise to compare the results of their experiment to a historical value?
What problems might be associated with this?

71. A researcher is investigating the actual weight of the nuts in packages of nuts that
have a stated weight of 250 grams. This researcher buys a sample of packages
and weighs the nuts in each package. (Assume that this sample can be thought
of as a random sample of bags of this type.) The researcher wishes to carry out
an appropriate hypothesis test and obtain a confdence interval for the true mean
weight of nuts in bags of this type. The following output summarizes the results of
the calculations.

One Sample t-test

data: nuts
t = 0.9966, df = 44, p-value = 0.3244
alternative hypothesis: true mean is not equal to 250
95 percent confidence interval:
248.4190 254.6741
sample estimates:
mean of x
251.5466

(a) What is the sample size?

(b) What is the p-value of the test of H0 : µ = 250 against the alternative Ha : µ >
250?
(c) What is the p-value of the test of H0 : µ = 250 against the alternative Ha : µ <
250?
(d) Give a proper interpretation of the results of the hypothesis test found in the
output.
(e) Of the following options, which one is the best interpretation of the confdence
interval found in the output?
i. We can be 95% confdent that the sample mean is within the interval
(248.4190, 254.6741).
Balka ISE 1.09 9.12. CHAPTER SUMMARY 151

ii. 95% of bags of nuts of this type weigh between 248.4190 grams and
254.6741 grams.
iii. We can be 95% confdent that the mean weight of all bags of nuts of this
type lies between 248.4190 grams and 254.6741 grams.
iv. In repeated sampling, 95% of the sample means will lie within the calcu-
lated interval.
v. In repeated sampling, 95% of the sampled bags of nuts will have a weight
between 248.4190 grams and 254.6741 grams.

72. In the United States, in order to label a food product as trans fat free, the product
must contain no more than 0.50 grams of trans fat per serving. (In Canada a lower
threshold of 0.2 g is used.) Suppose that a certain type of french fry in the United
States is labelled as trans fat free. A random sample of 150 servings yielded a
sample mean trans fat level of 0.538 g, with a sample standard deviation of 0.140
g.

Test the null hypothesis that the true mean trans fat content is 0.50 g, against the
alternative hypothesis that it is greater than 0.50 g.

(a) Which one of the following represents the appropriate hypotheses in symbols?
i. H0 : µ = 0.50, Ha : µ > 0.50
ii. H0 : X
¯ = 0.50, Ha : X¯ > 0.50
iii. H0 : µ = 0.50, Ha : µ < 0.50
iv. H0 : X¯ = 0.528, Ha : X¯ > 0.528
v. Both i and iv are appropriate.
(b) What is the value of the appropriate test statistic?
(c) What is the p-value of the test?
(d) Give an appropriate conclusion at the 0.05 signifcance level.
(e) Suppose the distribution of trans fat is strongly skewed (and therefore def-
nitely not normal). Would this hypothesis test still be informative?

73. A manufacturer of bolts is producing a type of bolt that is intended to have an

average weight of 82.90 grams. A random sample of 100 of these bolts found that
the weights were roughly normally distributed, with a sample mean weight of 84.14
grams and a sample standard deviation of 0.12 grams. Test the null hypothesis
that the population mean weight of this type of bolt is 82.9 grams, against the
alternative hypothesis that the population mean di˙ers from 82.9 grams.

(a) Is the appropriate test statistic a t statistic or a z statistic?

(b) What is the value of the appropriate test statistic?
(c) What is the p-value of the test?
(d) Give an appropriate conclusion at the 1% signifcance level.
(e) Suppose that the true mean weight of this type of bolt is in fact exactly equal
Balka ISE 1.09 9.12. CHAPTER SUMMARY 152

to 82.90 grams. In our hypothesis test, did we make a Type I error, a Type II
error, or neither error?

74. A manufacturer of automobiles has been crash testing cars for many years. In a
previous model, it was known from a large body of crash tests that the average
driver head injury rating for the crash-test dummies was 800 (units unknown, but
the larger the rating, the more damage). This was felt to be unacceptably high, so
a new design is implemented and new tests are conducted. Twenty-four cars are
crashed and the head injury rating for the drivers was measured. The sample mean
was found to be 750 and the sample standard deviation was found to be 50. Test the
null hypothesis that the population mean driver head injury for the new design is
800, against the alternative that the new design has a lower population mean head
injury rating. (Assume that the head injury ratings are normally distributed.)

(a) What are the appropriate hypotheses?

(b) What is the value of the appropriate test statistic?
(c) What is the p-value of the test?
(d) Give an appropriate conclusion at the 5% signifcance level.

75. A car manufacturer is investigating the fuel consumption of a new model of car. As
part of the study, 8 of these cars are driven for a certain amount of time under similar
real-world conditions and the fuel consumption (litres/100 km) is recorded. The
eight cars had a mean fuel consumption of 6.8 l/100k with a standard deviation of
0.32 l/100k. An older model of this car had a mean fuel consumption of 7.0 l/100k,
and the manufacturer wishes to test whether the true mean fuel consumption of
this new model di˙ers from 7.0 l/100k. Suppose that a normal quantile-quantile
plot showed that the fuel consumption values are roughly normally distributed, and
that we feel comfortable assuming normality.

(a) Test the null hypothesis that µ = 7, against a two-sided alternative hypothesis.
Give appropriate hypotheses, test statistic, p-value, and conclusion.
(b) What assumptions are necessary in order for this hypothesis test to be valid?
How might these assumptions be violated? What are the consequences if the
assumptions are violated?
Chapter 10

Inference for Two Means

J.B.’s strongly suggested exercises: 4, 6, 7, 9, 10, 11, 17, 23, 26, 27, 30, 31, 33

10.1 Introduction

10.2 The Sampling Distribution of the Di˙erence in Sam-

ple Means

1. Suppose we draw a random sample from Population 1, which has a mean of 15

and a standard deviation of 2, and we draw a random sample from Population 2,
which has a mean of 8 and a standard deviation of 6, and the two samples can be
considered independent.

(a) If n1 = n2 = 4, what is the mean of the sampling distribution of X¯1 − X¯2?

(b) If n1 = n2 = 4, what is the standard deviation of the sampling distribution of
X¯1 − X¯2?
(c) If n1 = n2 = 4, what is the shape of sampling distribution of X
¯1 − X¯2?
(d) If n1 = n2 = 400, what is the mean of the sampling distribution of X¯1 − X¯2?
(e) If n1 = n2 = 400, what is the standard deviation of the sampling distribution
of X ¯2?
¯1 − X
(f) If n1 = n2 = 400, what is the shape of sampling distribution of X
¯1 − X¯2?

2. Tom and Pete are two NFL prospects at the NFL combine, where players eligible
for the NFL draft show their skills in a variety of di˙erent exercises. Tom and Pete
are about to have three attempts at the 40 yard dash. Suppose that (theoretically)

153
 10.3. HYPOTHESIS TESTS AND CONFIDENCE INTERVALS FOR TWO INDEPENDENT SAMPLES (WHEN σ1 AND σ2 ARE KNOWN)

Balka ISE 1.09 154

Tom’s times in this event are approximately normally distributed with a mean of
4.612 seconds and a standard deviation of 0.048 seconds. Pete’s times are approxi-
mately normally distributed with a mean of 4.528 seconds and a standard deviation
of 0.044 seconds. Suppose it is reasonable to assume independence between runs.
(The runs may not be truly independent, but this assumption provides a reasonable
approximate model.)

(a) In their frst attempt, what is the probability that Tom’s time is greater than
Pete’s time?
(b) What is the probability that Tom’s average time in the three attempts is
greater than Pete’s average time in the three attempts?

10.3 Hypothesis Tests and Confdence Intervals for Two

Independent Samples (When σ1 and σ2 are known)

10.4 Hypothesis Tests and Confdence Intervals for µ1 −

µ2 (When σ1 and σ2 are unknown)

10.4.1 Pooled-Variance t Tests and Confdence Intervals

3. Table 10.1 illustrates the results of a two-sample study. The samples were drawn
independently from normally distributed populations.

Group 1 Group 2
Sample mean 8.8 17.2
Sample standard deviation 1.42 2.61
Sample size 10 5

Table 10.1: Results of a two-sample study.

For the pooled-variance t procedure, calculate:

(a) The point estimate of µ1 − µ2 .

(b) The pooled variance s2p .
(c) SE(X¯1 − X¯ 2 ).
(d) A 95% confdence interval for µ1 − µ2 .
(e) Test the null hypothesis that the population means are equal. Give the ap-
propriate hypotheses, standard error, value of the test statistic, p-value, and
conclusion at α = 0.05.
 10.4.
Balka ISE 1.09 HYPOTHESIS TESTS AND CONFIDENCE INTERVALS FOR µ1 − µ2 (WHEN σ1 AND σ2 ARE UNKNOWN) 155

4. Does a vitamin D supplement have an e˙ect on parathyroid hormone (PTH) levels

in the blood? An experiment1 investigated the e˙ect of a vitamin D supplement on
several biological factors in study participants. One of the variables was the change
in PTH in the blood.

In the experiment, 26 individuals were randomly assigned to one of two groups.

Each group consumed 240 mL of orange juice per day for 12 weeks. The orange
juice of the treatment group was fortifed with 1000 IU vitamin D3 , whereas the
control group’s orange juice had no vitamin D3 added. After 12 weeks, the change
in PTH level in the blood (pg/mL) was recorded. Table 10.4 illustrate the results.

Fortifed with vitamin D ¯ 1 = −9.0

X s1 = 37.5 n1 = 14
Not fortifed with vitamin D ¯ 2 = −1.6
X s2 = 34.6 n2 = 12

Table 10.2: Change in blood PTH levels after 12 weeks.

Suppose that normal quantile-quantile plots showed that the values were approxi-
mately normally distributed. Use the pooled-variance t procedure to answer the fol-
lowing questions. (Since the observations are approximately normally distributed,
and the sample standard deviations are similar, the pooled-variance t procedure is
a reasonable method of analysis.)

(a) Test the hypothesis that vitamin D has no e˙ect against a two-sided alter-
native. Give the appropriate hypotheses in words and symbols, value of the
test statistic, p-value, and conclusion. (Hint to ease the calculation burden:
s2p = 1310.417, SE(X ¯1 − X¯ 2 ) = 14.24088.)
(b) Calculate a 95% confdence interval for the di˙erence in population mean PTH
level. Give a proper interpretation of the interval.
(c) Give an overall summary of the results of the analysis.

10.4.2 Welch (Unpooled Variance) t Tests and Confdence Inter-

vals

5. Table 10.3 illustrates the results of a two-sample study. The samples were drawn
independently from normally distributed populations.

For the Welch procedure, calculate:

(a) The point estimate of µ1 − µ2 .

(b) SEW (X¯1 − X¯ 2 ).
1
Tangpricha et al. (2003). Fortifcation of orange juice with vitamin D: a novel approach for enhancing
vitamin D nutritional health. The American journal of clinical nutrition, 77(6): 1478-1483.
 10.4.
Balka ISE 1.09 HYPOTHESIS TESTS AND CONFIDENCE INTERVALS FOR µ1 − µ2 (WHEN σ1 AND σ2 ARE UNKNOWN) 156

Group 1 Group 2
Sample mean 8.8 17.2
Sample standard deviation 1.42 2.61
Sample size 10 5

Table 10.3: Results of a two-sample study.

(c) A 95% confdence interval for µ1 − µ2 . (Hint to ease the calculation burden:
DF = 5.221.)
(d) Test the null hypothesis that the population means are equal. Give the ap-
propriate hypotheses, standard error, value of the test statistic, p-value, and
a conclusion at α = 0.05. (Hint to ease the calculation burden: DF = 5.221.)

6. Consider again the study discussed in Question 4. (In this experiment, some in-
dividuals received orange juice fortifed with vitamin D, others received unfortifed
orange juice.) Does the vitamin D supplement a˙ect phosphorous levels in the
blood? The change in phosphorous levels over the course of the study are found in
the following table.

Fortifed with vitamin D ¯ 1 = −6.3

X s1 = 18.7 n1 = 14
Not fortifed with vitamin D ¯ 2 = −0.2
X s2 = 1.7 n2 = 12

Table 10.4: Change in blood phosphorous levels (mg/dL) after 12 weeks.

Note that the standard deviations are vastly di˙erent here (so much so that it makes
one wonder if there might be a typo in the article). Since the standard deviations
are vastly di˙erent, the pooled-variance procedure would start to break down and
would not be appropriate. The Welch procedure would be a better method of
analysis, so use the Welch procedure to answer the following questions.

(a) What is the value of SEW (X

¯1 − X
¯ 2 )?

(b) If we wish to test the null hypothesis that, on average, vitamin D has no e˙ect
on phosphorous levels, what is the value of the appropriate test statistic?

(c) The R output for the Welch procedure is:

Balka ISE 1.09 10.5. PAIRED-DIFFERENCE PROCEDURES 157

Welch Two Sample t-test

data: vitaminD and novitaminD
t = -1.2156, df = 13.215, p-value = 0.2454
alternative hypothesis: true difference in means is not equal to 0
95 percent confidence interval:
-16.923691 4.723691
sample estimates:
mean of x mean of y
-6.3 -0.2

Give a summary of the results of the analysis.

10.4.3 Guidelines for Choosing the Appropriate Two-Sample t Pro-

cedure

7. What factors infuence the choice between the Welch procedure and the pooled-
variance t procedure? When is the Welch procedure a better choice? When is the
pooled-variance t procedure a better choice?

8. Suppose we are interested in assessing the e˙ect of two fuel additives on fuel ef-
fciency. We intend to use a t procedure to carry out an appropriate hypothesis
test. Under which of the following situations would it be most appropriate to use
Welch’s approximation instead of the pooled-variance procedure?

(a) The sample sizes of the two groups are similar, and the sample standard de-
viations are similar.
(b) The sample sizes of the two groups are similar, and the sample standard de-
viations are very di˙erent.
(c) The sample sizes of the two groups are very di˙erent, and the sample standard
deviations are similar.
(d) The sample sizes of the two groups are very di˙erent, and the sample standard
deviations are very di˙erent.

10.5 Paired-Di˙erence Procedures

10.5.1 Paired-Di˙erence t Tests and Confdence Intervals

9. O˙erman et al. (2009) conducted an experiment on pigs to investigate the e˙ect of

an antivenom after an injection of rattlesnake venom. In one aspect of the study,
the researchers investigated the change in volume of the right hind leg before and
Balka ISE 1.09 10.5. PAIRED-DIFFERENCE PROCEDURES 158

after being subjected to a dose of venom and treatment with an antivenom. The
volume of the right hind leg was measured in 9 pigs before being injected with the
venom, then the pigs were injected with a dose of venom and treated intravenously
with an antivenom, and after 8 hours the volume of the leg was measured again.
The results are illustrated in Table 10.5. (The volume was measured using a water
displacement method; the units are mL.)

Before After Di˙erence

685 935
545 700
480 770
475 640
680 800
685 955
590 780
600 790
630 830

Table 10.5: Volume (mL) of the right hind leg before and 8 hours after injection with a
dose of rattlesnake venom and an antivenom.

(a) Suppose we are interested in estimating the true mean amount of swelling for
pigs of this type under the conditions of this experiment. Would we use a
paired-di˙erence procedure or an independent sample procedure? Justify your
response.
(b) What are the 9 di˙erences? (Take the di˙erences as After − Before.)
(c) What is the sample mean di˙erence? What is the standard deviation of the
di˙erences?
(d) What is the standard error of the sample mean di˙erence?
(e) The boxplot and normal quantile-quantile plot of the di˙erences are given in
Figure 10.1. Do these plots give any indication that the t procedures should
not be used?
(f) Construct a 95% confdence interval for the true mean amount of swelling under
the conditions of this experiment. Give an interpretation of the confdence
interval in the context of the problem at hand.
(g) Carry out a test of the null hypothesis that there is no swelling on average
(after 8 hours), against the appropriate one-sided alternative hypothesis. Give
the hypotheses in words and symbols, value of the test statistic, p-value, and
conclusion.

10. Porter et al. (2010) investigated the e˙ect of a brief training program on the ability
of health care professionals to detect deception. In part of the study, 26 health
care workers were shown videos in which individuals sometimes showed a genuine
Balka ISE 1.09 10.5. PAIRED-DIFFERENCE PROCEDURES 159

Change in Volume (ml)

250

250
Sample Quantiles
150

200
50

150
0

-1.5 -1.0 -0.5 0.0 0.5 1.0 1.5

Theoretical Quantiles

(a) The line represents the hypothesized value of (b) Normal quantile-quantile plot.
0.

Figure 10.1: Changes in volume (mL) of the right hind leg before and 8 hours after
injection with a dose of rattlesnake venom and an antivenom.

smile, and sometimes showed a fake smile. Individuals completed this task before
and after completing a 3 hour session designed to help them recognize deception.
The participants’ responses were evaluated and they were given a discrimination
accuracy score (high positive scores indicate a person is correctly discriminating
between a genuine smile and a fake smile, values near 0 indicate the individual is
not discriminating between a genuine smile and a fake smile, and negative values
indicate that the individual is misclassifying genuine smiles as fake and fake smiles
as genuine). The results of the study are given in Table 10.6.

Mean SD
Before 0.192 0.770
After 0.810 1.173
Di˙erence (After − Before) 0.618 1.412

Table 10.6: Discrimination accuracy scores for 26 health care professionals before and
after a training program.

The researchers were interested in testing whether the training program had an
e˙ect, and in estimating the size of the e˙ect with a confdence interval.

(a) What should be done before using software to carry out the t procedures?
(b) When the calculations are carried out in R, the default output is:
Balka ISE 1.09 10.6. INVESTIGATING THE NORMALITY ASSUMPTION 160

One Sample t-test

data: expressions
t = 2.2317, df = 25, p-value = 0.0348
alternative hypothesis: true mean is not equal to 0
95 percent confidence interval:
0.04768086 1.18831914
sample estimates:
mean of x
0.618
The p-value given in the output iss for a two-sided alternative hypothesis. But
suppose that the researchers felt a one-sided alternative hypothesis was more
appropriate, since they felt that, if anything, the training session would im-
prove a participant’s ability to discriminate between genuine and fakes smiles.
What would be the appropriate one-sided alternative? What would the p-value
be for this test?
(c) Give a summary of the results of the analysis. Use a signifcance level of 0.05.

10.6 Investigating the Normality Assumption

10.7 Chapter Exercises

10.7.1 Basic Calculations

10.7.2 Concepts

11. In words, what is the meaning of SE(X

¯1 − X
¯ 2 )?

12. Would it make sense to test the hypothesis H0 : X ¯ 2 ? Why or why not?
¯1 = X

13. A 95% confdence interval for µ1 − µ2 is found to be (2, 28).

(a) Give an example of a null hypothesis that would be rejected at α = 0.05.

(b) Give an example of a null hypothesis that would not be rejected at α = 0.05.

14. Consider the following output for a two-sample problem.

Balka ISE 1.09 10.7. CHAPTER EXERCISES 161

Two Sample t-test

t = -0.5692, df = 18, p-value = 0.5763
alternative hypothesis: true difference in means is not equal to 0
95 percent confidence interval:
-0.9817025 0.5631646
sample estimates:
mean of x mean of y
-0.12739020 0.08187872

¯ 1 does not equal X

Does this output yield strong evidence that X ¯2?

15. Suppose we wish to test the null hypothesis that µ1 − µ2 = 0 against a two-sided
alternative hypothesis.

(a) All else being equal, what will happen to the power of the test as the true
di˙erence µ1 − µ2 gets closer to 0?
(b) All else being equal, what will happen to the power of the test as the sample
sizes increase?

16. Consider a two-sample t test of the null hypothesis of equal population means
against a two-sided alternative.

(a) Under what conditions would the test statistic be equal to 0?

(b) Under what conditions would the p-value be equal to 1?
(c) Under what conditions would the p-value be equal to 0?

17. Consider the three boxplots in Figure 10.2, which represent samples of size 40 from
3 di˙erent populations. Consider the following 3 null hypotheses (with two-sided
5
0
-5

A B C

Figure 10.2

alternatives in all cases).

I. H0 : µA = µB
II. H0 : µA = µC
Balka ISE 1.09 10.7. CHAPTER EXERCISES 162

III. H0 : µB = µC

(a) Which test would result in the smallest p-value?

(b) Which test would result in the largest p-value?

18. Consider the following output for a two-sample inference procedure, and the corre-
sponding boxplots in Figure 10.3.

A B

Figure 10.3

Two Sample t-test

t = -1.0655, df = 43, p-value = 0.2926
alternative hypothesis: true difference in means is not equal to 0
95 percent confidence interval:
-1.6882973 0.5209971
sample estimates:
mean of x mean of y
4.367041 4.950691

(a) Do the boxplots give any indication that this t procedure should not be used?
(b) Is there strong evidence that the populations have di˙erent means?
(c) Based on the confdence interval in the output, is there strong evidence that
µ1 − µ2 6= 10?
(d) What is the sum of the two sample sizes?

19. Suppose we draw two independent samples of sizes n1 = 100 and n2 = 50, and wish
to test the null hypothesis that the population means are equal. The output for
the two procedures (Welch and the pooled-variance t) are:

Output 1:
Balka ISE 1.09 10.7. CHAPTER EXERCISES 163

data: sample1 and sample2

t = -1.3734, df = 98.659, p-value = 0.1727
alternative hypothesis: true difference in means is not equal to 0
95 percent confidence interval:
-0.5677384 0.1032975
sample estimates:
mean of x mean of y
-0.1298456 0.1023749

Output 2:

data: sample1 and sample2

t = -1.3705, df = 148, p-value = 0.1726
alternative hypothesis: true difference in means is not equal to 0
95 percent confidence interval:
-0.5670558 0.1026149
sample estimates:
mean of x mean of y
-0.1298456 0.1023749

(a) Which output is from the pooled-variance procedure? Which output is from
the Welch procedure?
(b) Which procedure would be more appropriate here?

20. Forty patients with high blood pressure volunteer for a study. The participants are
given an injection of a placebo (a saline solution with no pharmacological e˙ect),
and an hour later their drop in blood pressure is measured. Two days later the same
40 patients are given an injection of an experimental new drug, and an hour later
the drop in blood pressure is recorded. Researchers want to investigate the e˙ect
of the drug on blood pressure. They wish to compare the drop in blood pressure
after the injection of the drug to the drop in blood pressure after injection of the
placebo.

(a) What t procedure would be the most appropriate procedure to use here (Welch’s,
pooled-variance, or paired di˙erence)? Suppose that it is reasonable to assume
normality where necessary.
(b) Suppose we observe a signifcant di˙erence between the drop in blood pressure
observed after the placebo injection, and the drop in blood pressure after the
drug injection. Other than a possible e˙ect of the drug, are there any features
of this study design that may possibly be the cause of this di˙erence?

21. A research experiment was designed to assess the e˙ectiveness of two memory train-
ing programs. The experiment involved 25 sets of identical twins. One twin in each
pair was randomly assigned to Program A, the other to Program B. Upon comple-
tion of the program, everyone took a memory test and obtained a score between 0
and 100. If we want to carry out a hypothesis test to see if one of the programs
Balka ISE 1.09 10.7. CHAPTER EXERCISES 164

tends to result in a better score on the memory test, what is the most appropriate
test to use? Assume normality where necessary.

22. (Challenge! We haven’t worked through a problem like this—you need to think
it through.) Suppose we wish to draw independent random samples from two
populations. Suppose that it is known that σ1 = σ2 = 3, but the population
means are unknown. We intend to draw equal sample sizes n1 = n2 = n from both
populations. If we wish to estimate µ1 − µ2 within 0.5 with 95% confdence, what
sample size would be required?

23. Test your conceptual understanding: Which of the following statements are true?
You should be able to explain why a statement is true or why a statement is false.

(a) When we use the pooled-variance t procedure, it is because we know the pop-
ulations have the same variance.
(b) The pooled-variance t procedure works well, even when the population vari-
ances are a little di˙erent. This is especially true if the sample sizes are similar.
(c) It would be most appropriate to use the Welch procedure instead of the pooled-
variance t procedure if the sample variances are very di˙erent and the sample
sizes are very di˙erent.
(d) If the conclusions from the Welch procedure and the pooled-variance t proce-
dure are very similar, then it does not matter much which procedure is used.
(e) The Welch procedure is an exact procedure, as long as X ¯1 = X¯2.

24. Test your conceptual understanding: Which of the following statements are true?
You should be able to explain why a statement is true or why a statement is false.

(a) If µ1 = µ2 , the sampling distribution of X ¯ 2 is approximately symmetric

¯1 − X
about 0 for large sample sizes.
(b) If X
¯1 = X¯ 2 , the sampling distribution of µ1 − µ2 is approximately symmetric
about 0 for large sample sizes.
(c) SE(X ¯1 − X ¯ 2 ) is the true standard deviation of the sampling distribution of
X1 − X2 .
¯ ¯
(d) The pooled-variance t procedures work well, even when the variances for the
two populations are very di˙erent, as long as the sample sizes are very di˙erent
as well.
(e) Suppose we about to test the null hypothesis µ1 = µ2 against a two-sided
alternative. All else being equal, the greater the di˙erence between µ1 and µ2 ,
the greater the power of the test.

25. Test your conceptual understanding: Which of the following statements are true?
You should be able to explain why a statement is true or why a statement is false.

(a) Suppose we are constructing a confdence interval for µ1 − µ2 . All else being
Balka ISE 1.09 10.7. CHAPTER EXERCISES 165

equal, the greater the di˙erence between µ1 and µ2 , the wider the interval.
(b) Suppose we are constructing a confdence interval for µ1 − µ2 . All else being
equal, the greater the di˙erence between X ¯ 1 and X
¯ 2 , the wider the interval.
(c) Suppose we are constructing a confdence interval for µ1 − µ2 . All else being
equal, the greater the sample sizes, the narrower the interval.
(d) Suppose we wish to test H0 : µ1 = µ2 . We obtain random samples from the
respective populations, run the appropriate test, and fnd that the p-value is
0.00000032. We can be very confdent that our results have important practical
implications.
(e) If we test H0 : µ1 = µ2 against a two-sided alternative and fnd a p-value of
0.32, then we know that µ1 = µ2 .

10.7.3 Applications

26. A study2 investigated various aspects of tandem running in ants. Tandem running
is a form of recruitment in which one ant with knowledge of the location of a food
source or new nest site leads another ant to that location. (Optional: Watch an ex-
ample of tandem running here: (1:53) (http://www.youtube.com/watch?v=X2C7Sy2oPik))
Tandem running can be thought of as a form of teaching and learning.

The study investigated various factors associated with tandem running, and inves-
tigated whether the age of the ant and the experience of the ant had any e˙ect on
tandem running. Ants (Temnothorax albipennis) were categorized into 4 categories:
Young and Inexperienced (YI), Young and Experienced (YE), Old and Inexperi-
enced (OI), Old and Experienced (OE). One aspect of the study investigated the
speed of tandem running when leading a YI ant. The mean speed of the run was
recorded for 51 tandem runs in which the leader was YE, and 15 pairs of runs in
which the leader was OE. The researchers were interested in a possible di˙erence
in the mean speed of tandem running in these two situations.

Figures 10.4 and 10.5 and Table 10.7 illustrate the data.

OE ¯ 1 = 1.14
X s1 = 0.37 n1 = 15
YE ¯ 2 = 1.80
X s2 = 0.60 n2 = 51

Table 10.7: Means and standard deviations for the tandem running study.

(a) Do the plots give any indication that the t procedure should not be used?
2
Franklin et al. (2012). Do ants need to be old and experienced to teach? The Journal of Experimental
Biology, 215:1287–1292 The data used here is simulated data based on their Figure 2, with similar results
and conclusions.
Balka ISE 1.09 10.7. CHAPTER EXERCISES 166

2.5
2.0
Speed (mm/s)
1.5
1.0
0.5

OE YE

Figure 10.4: Speed of tandem running for ant pairs led by an old experienced (OE) ant,
and ant paris led by a young experienced (YE) ant.

Normal Q-Q Plot Normal Q-Q Plot

1.6

2.5
0.8 1.0 1.2 1.4
Sample Quantiles

Sample Quantiles
1.0 1.5 2.0
0.6
0.4

0.5

-1 0 1 -2 -1 0 1 2
Theoretical Quantiles Theoretical Quantiles

(a) OE ants. (b) YE ants.

Figure 10.5: Normal QQ plots for OE and YE ants.

(b) Which version of the t procedure (pooled or unpooled) is more appropriate

here?
(c) Using the Welch procedure, construct a 95% confdence interval for the di˙er-
ence in true mean tandem running speed between OE and YE ants. (Hint to
ease the calculation burden: SEW (X ¯1 − X¯ 2 ) = 0.1272, DF = 37.714.)
(d) The researchers were interested in investigating a possible di˙erence in the
true mean running speeds for the two types of ant. In words and symbols,
what are the hypotheses of the appropriate hypothesis test?
(e) What is the value of the appropriate test statistic?
(f) What is the p-value of the test? Is there strong evidence against the null
hypothesis?
(g) The output from the statistical software R for the Welch procedure is:
Balka ISE 1.09 10.7. CHAPTER EXERCISES 167

Welch Two Sample t-test

data: OE and YE
t = -5.1878, df = 37.714, p-value = 7.516e-06
alternative hypothesis: true difference in means is not equal to 0
95 percent confidence interval:
-0.9176121 -0.4023879
sample estimates:
mean of x mean of y
1.14 1.80

Give a summary of the results of the analysis (including the results of the
hypothesis test and the confdence interval). To what population do your
conclusions apply?

27. A study3 investigated several aspects of the ratio of the lengths of the index fnger
to the ring fnger in women who visited a sexual health clinic in Manchester, UK.
(The ratio of the lengths of these fngers is called the 2D:4D ratio.) The distribution
of the 2D:4D ratio depends on several factors, including the ethnic background of
the individuals. The following table summarizes the results of the measurements of
the 2D:4D ratio on the left hand for white and black women in the study.

White women ¯ 1 = 0.994

X s1 = 0.035 n1 = 246
Black women ¯
X2 = 0.963 s2 = 0.034 n2 = 46

(a) What plots should be created before carrying out any inference procedures?
(b) Which version of the t procedure (pooled or unpooled) is more appropriate
here?
(c) Using the pooled-variance t procedure, construct a 95% confdence interval for
the di˙erence in the means of the 2D:4D ratio. (Hint to ease the calculation
burden: SE(X ¯1 − X¯ 2 ) = 0.005598.)
(d) Suppose we wish to test whether there is strong evidence of a di˙erence be-
tween the true means of the 2D:4D ratio for white and black women. In words
and symbols, what are the hypotheses of the appropriate hypothesis test?
(e) What is the value of the appropriate test statistic?
(f) What is the p-value of the test? Is there strong evidence against the null
hypothesis?
(g) The output from the statistical software R for the pooled-variance procedure
is:
3
Brabin et al. (2008). The second to fourth digit ratio (2d:4d) in women with and without human
papillomavirus and cervical dysplasia. American Journal of Human Biology, 20:337–341.
Balka ISE 1.09 10.7. CHAPTER EXERCISES 168

Two Sample t-test

data: white and black
t = 5.538, df = 290, p-value = 6.859e-08
alternative hypothesis: true difference in means is not equal to 0
95 percent confidence interval:
0.01998280 0.04201720
sample estimates:
mean of x mean of y
0.994 0.963
Give a summary of the results of the analysis (including the results of the
hypothesis test and the confdence interval). To what population do your
conclusions apply?

28. Researchers used an experiment to investigate whether prolonged food restriction

might lead to a relapse of drug abuse.4 Rats were randomly assigned to a sated
(well fed) group or a prolonged food restriction group (the rats were under food
restriction for 10 days, and their weight fell to approximately 80% of the weight of
the sated rats). Before the food deprivation period, all rats were trained to use a
lever to self-administer heroin. Access to heroin was removed for all rats during the
10 day food deprivation period. Rats were then given access to the heroin lever, and
the number of active lever presses in a 3 hour period was measured. The following
table summarizes the results.

Food restriction ¯ 1 = 48.9

X s1 = 25.4 n1 = 8
Sated (control) ¯ 2 = 23.8
X s2 = 7.9 n2 = 7

(a) What plots should be created before carrying out any statistical inference?
(b) Plots of the data (not shown) show some right skewness in both groups, and
so the use of the t procedures is a bit dubious. But if the skewness is similar in
both groups, the t procedure may still perform reasonably well. Assuming that
we choose to use the t procedures, which version of the t procedure (pooled or
unpooled) is more appropriate here?
(c) Using the Welch (unpooled variance) t procedure, construct a 95% confdence
interval for the di˙erence in the means between the groups. (Hint to ease the
calculation burden: SEW (X ¯1 − X ¯ 2 ) = 9.4637, DF = 8.512.)
(d) Suppose we wish to test whether there is strong evidence of a treatment e˙ect
(a di˙erence between the true means of the two groups). In words and symbols,
what are the hypotheses of the appropriate hypothesis test?
(e) What is the value of the appropriate test statistic?
(f) What is the p-value of the test? Is there strong evidence against the null
hypothesis?
4
Shalev (2011). Chronic food restriction augments the reinstatement of extinguished heroin-seeking
behavior in rats. Addiction Biology, 17:691–693.
Balka ISE 1.09 10.7. CHAPTER EXERCISES 169

(g) The output from the statistical software R for the pooled-variance procedure
is:
Welch Two Sample t-test
data: Restricted and Sated
t = 2.6523, df = 8.512, p-value = 0.02764
alternative hypothesis: true difference in means is not equal to 0
95 percent confidence interval:
3.50315 46.69685
sample estimates:
mean of x mean of y
48.9 23.8
Give a summary of the results of the analysis (including the results of the
hypothesis test and the confdence interval). To what population do your
conclusions apply?

29. A study5 investigated several physical characteristics of the ears of Italian Cau-
casians. In one part of the study, 3D symmetry (a measure of the symmetry be-
tween the left and right ears) was measured on men and women between 31 and 40
years of age.

Men ¯ 1 = 95.71
X s1 = 1.50 n1 = 66
Women ¯
X2 = 95.41 s2 = 1.64 n2 = 28

(a) What plots should be created before carrying out any inference procedures?
(b) Which version of the t procedure (pooled or unpooled) is more appropriate
here?
(c) Using the pooled-variance t procedure, construct a 95% confdence interval
for the di˙erence in the means of the 3D symmetry index. (Hint to ease the
calculation burden: SE(X ¯1 − X
¯ 2 ) = 0.3479.)
(d) Suppose we wish to test whether there is strong evidence of a di˙erence be-
tween the true mean 3D symmetry in men and women. In words and symbols,
what are the hypotheses of the appropriate hypothesis test?
(e) What is the value of the appropriate test statistic?
(f) What is the p-value of the test? Is there strong evidence against the null
hypothesis?
(g) The output from the statistical software R for the pooled-variance procedure
is:
5
Sforza et al. (2009). Age- and sex-related changes in the normal human ear. Forensic Science Interna-
tional, 187:110.e1–110.e7.
Balka ISE 1.09 10.7. CHAPTER EXERCISES 170

Two Sample t-test

data: men and women
t = 0.8624, df = 92, p-value = 0.3907
alternative hypothesis: true difference in means is not equal to 0
95 percent confidence interval:
-0.3908909 0.9908909
sample estimates:
mean of x mean of y
95.71 95.41
Give a summary of the results of the analysis (including the results of the
hypothesis test and the confdence interval). To what population do your
conclusions apply?

30. A study6 investigated several characteristics of psychopathic and nonpsychopathic

male patients in a Dutch inpatient psychiatric treatment centre. In one part of the
study, the BEST index (a measure of risk behaviours of psychiatric patients) was
measured after several months of treatment. The higher the score on the BEST
index, the better (less risky) the patient scored. The following table summarizes
the results.

Nonpsychopaths ¯ 1 = 256.12
X s1 = 33.53 n1 = 47
Psychopaths ¯ 2 = 245.82
X s2 = 37.55 n2 = 27

(a) What plots should be created before carrying out any inference procedures?
(b) Which version of the t procedure (pooled or unpooled) is more appropriate
here?
(c) Using the pooled-variance t procedure, construct a 95% confdence interval for
the di˙erence in the means of the BEST index. (Hint to ease the calculation
burden: SE(X ¯1 − X¯ 2 ) = 8.4603.)
(d) Suppose we wish to test whether there is strong evidence of a di˙erence be-
tween the true means of the BEST index for psychopaths and nonpsychopaths.
In words and symbols, what are the hypotheses of the appropriate hypothesis
test?
(e) What is the value of the appropriate test statistic?
(f) What is the p-value of the test? Is there strong evidence against the null
hypothesis?
(g) The output from the statistical software R for the pooled-variance procedure
is:
6
Chakhssi et al. (2010). Change during forensic treatment in psychopathic versus nonpsychopathic
o˙enders. 21:660–682.
Balka ISE 1.09 10.7. CHAPTER EXERCISES 171

Two Sample t-test

data: psychopaths and non psychopaths
t = 1.2174, df = 72, p-value = 0.2274
alternative hypothesis: true difference in means is not equal to 0
95 percent confidence interval:
-6.565314 27.165314
sample estimates:
mean of x mean of y
256.12 245.82
Give a summary of the results of the analysis (including the results of the
hypothesis test and the confdence interval). To what population do your
conclusions apply?

31. What e˙ect does a shock with a Taser have on the components of blood? Jauchem
et al. (2013) investigated this in an experiment on 11 wild boar (Sus scrofa) by
measuring characteristics of blood before and after a Taser shock. One measured
characteristic was the mean corpuscular volume (the average volume of red blood
cells, measured in femtolitres (10−15 L)).

Mean SD
Before 54.5 2.4
After 57.8 2.6
Di˙erence (After − Before) 3.3 3.7

Table 10.8: Summary statistics of the mean corpuscular volume (fL) for 11 wild boar
before and after a Taser shock.

Suppose that the changes in mean corpuscular volume are approximately normally
distributed.

(a) Construct a 95% confdence interval for the true mean change in mean corpus-
cular volume, and give an interpretation of the interval.
(b) Carry out a test of the null hypothesis that, on average, there is no change
in mean corpuscular volume, against a two-sided alternative hypothesis. Give
the hypotheses in words and symbols, value of the test statistic, p-value, and
conclusion.

10.7.4 Extra Practice Questions

32. In an observational study involving 42 new mothers, the women were classifed
according to whether or not they had used marijuana in their pregnancy. The
point of interest was comparing the birth weight of the babies born to the two
groups.
Balka ISE 1.09 10.7. CHAPTER EXERCISES 172

Did not use marijuana ¯ 1 = 3468

X s1 = 680 n1 = 32
Used marijuana ¯ 2 = 3200
X s2 = 610 n2 = 10

The researchers wanted to use a reasonable statistical analysis to investigate if

there is a di˙erence in mean birth weight between the two groups of mothers. The
researchers felt it was reasonable to assume normality and to assume the two popu-
lation variances are equal. Using the appropriate procedure under these conditions,
they researchers found an output of:

Two Sample t-test

data: marijuana and nomarijuana
t = 1.1126, df = 40, p-value = 0.2725
alternative hypothesis: true difference in means is not equal to 0
90 percent confidence interval:
****** ******
sample estimates:
mean of x mean of y
3468 3200

(a) The 90% confdence interval has been omitted from the output. What is the
appropriate interval? (Hint to ease the calculation burden: SE(X
¯1 − X
¯2) =
240.8803.)

(b) Which one of the following is the most appropriate conclusion?

i. There is very strong evidence that the population mean weight of babies
born to mothers who used marijuana is equal to that of babies born to
mothers who did not use marijuana.
ii. There is not strong evidence of a di˙erence in population mean birth
weight between the two groups.
iii. There is strong evidence that the population mean birth weights for the
two groups are equal.
iv. There is not strong evidence that the sample mean birth weights are dif-
ferent.
v. There is very strong evidence that marijuana causes a reduction in birth
weight.

33. An experiment was designed to investigate the e˙ect of exercise on the size of
tumours in rats. Thirty rats were injected with cancerous cells. These rats were
then randomly assigned to two groups. Ten mice were kept in cages with exercise
wheels, and 20 were kept in cages with no wheels. After six weeks, the diameter of
the tumour (cm) was recorded. The following table summarizes the results:
Balka ISE 1.09 10.7. CHAPTER EXERCISES 173

Exercise Wheels No Wheels

Sample mean 1.2 1.5
Sample variance 0.08 0.12

(a) Is this an observational study or an experiment?

(b) The researcher wanted to test whether the exercise wheels reduced the size
of tumours in rats. Carry out the appropriate hypothesis test. Give appro-
priate hypotheses, the value of the test statistic, p-value, and an appropriate
conclusion. (Assume equal population variances, and that the tumour sizes
are normally distributed.) (Hint to ease the calculation burden: s2p = 0.10714,
SE(X ¯1 − X
¯ 2 ) = 0.12677)
(c) Suppose we found a p-value of .00000067 in the previous question. Would this
give strong evidence of a causal link between exercise and size of tumour in
this type of experiment?
(d) Calculate a 90% confdence interval for the di˙erence in mean tumour size (for
the pooled variance case).
(e) Now, not assuming equal population variances, carry out the appropriate test.
Give appropriate hypotheses, value of the test statistic, p-value, and an ap-
propriate conclusion. (Hint to ease the calculation burden: DF = 21.764.)
(f) Calculate a 90% confdence interval for the di˙erence in mean tumour size.
(Not assuming equal population variances.) (Hint to ease the calculation bur-
den: DF = 21.764.)
(g) Which procedure (pooled-variance or Welch) is more appropriate in this case?

34. Researchers investigated the total cholesterol levels in the blood of male and female
students at a large university. Total cholesterol (mg/dl) was measured on 26 male
and 22 female student volunteers, with the following results.

Males Females
Sample mean 171.4 173.8
Sample standard deviation 32.9 34.1
Sample size 26 22

Although the people in the study are volunteers, for the purposes of these questions
assume they can be thought of as random samples from the populations.

(a) The researchers wanted to test the null hypothesis that the population mean
total cholesterol level is the same for both males and females, against a two-
sided alternative hypothesis. The following output represents the results of
the pooled variance two-sample t procedure.
Balka ISE 1.09 10.7. CHAPTER EXERCISES 174

Two Sample t-test

data: males and females
t = -0.2477, df = 46, p-value = 0.8055
alternative hypothesis: true difference in means is not equal to 0
95 percent confidence interval:
-21.90658 17.10658
sample estimates:
mean of x mean of y
171.4 173.8

Of the following options, which one is the most appropriate conclusion at

α = .05?

i. There is strong evidence of a di˙erence in population means.

ii. There is signifcant evidence that females have a higher population mean
total cholesterol level.
iii. The observed di˙erence in sample mean cholesterol levels is signifcant.
iv. There is not signifcant evidence that the sample mean cholesterol levels
are di˙erent.
v. There is not signifcant evidence that the population mean cholesterol
levels are di˙erent.

(b) Give an appropriate interpretation of the confdence interval found in the out-
put.

35. In an investigation into the e˙ectiveness of two processes that reduce contaminants
in used motor oil, 18 batches of used motor oil were randomly assigned to the
two processes. 12 were randomly assigned to process A, and 6 to process B. The
contaminant level after processing was measured for both processes. The results
are given in Table 10.9.

A B
Sample size 12 6
Sample variance 16 9
Sample mean 63 57

Table 10.9: Summary of information for the motor oil question.

Suppose that the normality assumption of the t procedures is in fact reasonable, and
that the population variance of the contaminants is the same for both processing
methods.

(a) Calculate a 95% confdence interval for the di˙erence in the population mean
contaminant levels (µA − µB ). Give an appropriate interpretation of the in-
terval.
Balka ISE 1.09 10.7. CHAPTER EXERCISES 175

(b) Carry out a test of the null hypothesis that the true mean contaminant level
is the same for both processes, against the alternative hypothesis that it is
di˙erent. Give appropriate hypotheses in words and symbols, value of the test
statistic, and p-value.
(c) Summarize the results of the analysis.

36. Many studies have investigated a connection between “fear of negative evaluation”
and bulimia. Suppose that researchers at a large university are interested in carry-
ing out their own investigation. Eleven female students with bulimia completed a
questionnaire and were assigned a “fear of negative evaluation” score, with a result-
ing sample mean of 19.7. Fourteen female students with normal eating habits were
given the same questionnaire, with a resulting sample mean “fear of negative evalu-
ation” score of 14.9. The researchers ran a pooled-variance two-sample t procedure
on the data, with the following results.

Two Sample t-test

data: bulimic and notbulimic
t = 2.5621, df = 23, p-value = 0.01742
alternative hypothesis: true difference in means is not equal to 0
95 percent confidence interval:
0.9380235 8.8022362
sample estimates:
mean of x mean of y
19.72727 14.85714

Assume that the students can be thought of as random samples from the populations
of female students at the university, and the other assumptions of the pooled-
variance t-procedures are reasonable in this case.

(a) In words and symbols, what are the hypotheses of the t test given in the
output?
(b) Give an appropriate conclusion at α = .05.
(c) Give a proper interpretation of the confdence interval found in the output.

37. As part of a production process, a company needs a certain type of resistor to have
a resistance of 12 ohms. The company obtains resistors from 2 di˙erent suppliers
and tests their resistance. The results are given in the table below. Although it
would not be perfectly justifed in this case, assume the resistors can be thought of
an independent random samples from their respective populations.

Manufacturer Sample size Sample mean Sample standard deviation

Manufacturer A 12 11.60 0.65
Manufacturer B 16 12.20 0.42

The researchers want to investigate several properties of these resistors. As part of

Balka ISE 1.09 10.7. CHAPTER EXERCISES 176

12.8
12.5

12.6
12.0

12.4
Sample Quantiles

Sample Quantiles

12.2
11.5

12.0
11.8
11.0

11.6
-1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 -2 -1 0 1 2

Theoretical Quantiles Theoretical Quantiles

(a) Manufacturer A (b) Manufacturer B

Figure 10.6: Normal quantile-quantile plots of the resistance.

their frst investigation, they want to estimate the di˙erence in the mean resistance
between the two types of resistor, and test the null hypothesis that the resistors
have the same resistance on average. They wish to use the t procedures, but realize
that they should frst investigate the normality assumption. Figure 10.6 illustrates
the normal quantile-quantile plots in this scenario.

(a) Do these plots give any indication that the normality assumption is violated?
(b) As part of their analysis into the di˙erences between the resistors, the re-
searchers want to test the null hypothesis that the population mean resistance
is the same for both types of resistor, against the alternative hypothesis that
the mean resistance is di˙erent. They also wish to calculate a 95% confdence
interval for the di˙erence in the mean resistance. They decide to use the
pooled-variance t procedure. The Welch procedure would have also been
a reasonable choice (the output for the Welch procedure is included farther
below). What is the value of the pooled sample variance (s2p )?
(c) What are the appropriate hypotheses in words and symbols?
(d) What is the value of the appropriate t statistic?
(e) Give an appropriate conclusion at the 5% signifcance level.
(f) What is a 95% confdence interval calculated using the pooled-variance t pro-
cedure?
(g) The following output represents the results of the Welch procedure on the data
above, using the alternative hypothesis that the population mean resistance
for the two types of resistors is di˙erent.
Welch Two Sample t-test
data: resistors
t = -2.7904, df = 17.695, p-value = 0.01222
alternative hypothesis: true difference in means is not equal to 0
95 percent confidence interval:
-1.0522974 -0.1477026
Balka ISE 1.09 10.7. CHAPTER EXERCISES 177

sample estimates:
mean of x mean of y
11.6 12.2
Based on the confdence interval in the output, what can be said about possible
di˙erences in the mean resistance?
(h) Refer again to the output from the Welch procedure. Note that the output
gives the p-value for a two-sided alternative. Had the researchers felt that
Ha : µA < µB was the more appropriate alternative (before looking at the
data), what would the p-value of that test be?

38. A company is interested in investigating properties of a new aluminum alloy that is

produced by a new experimental process. They wish to compare the yield strength
of this new alloy to the yield strength of a standard alloy that is commonly used.
The following table gives a summary of the yield strength (MPa) for independent
samples of each type of alloy.

Alloy type Sample size Sample mean Sample standard deviation

New alloy 80 664.2 23.3
Standard alloy 75 624.3 19.8

Suppose it is reasonable to assume that the alloy strengths are approximately nor-
mally distributed, and we wish to carry out inference procedures to investigate
possible di˙erences between the alloys.

There is little di˙erence between the sample standard deviations, so the use of the
pooled-variance t procedure is reasonable here. Use the pooled-variance t procedure
to answer the following questions.

(a) Calculate a 95% confdence interval for the di˙erence in true mean yield
strength. Give a proper interpretation of the interval. (Hint to ease the cal-
culation burden: s2p = 469.9299, SE(X ¯1 − X¯ 2 ) = 3.4842.)
(b) Carry out a test of the null hypothesis that the population means are equal,
against a two-sided alternative hypothesis. Give appropriate hypotheses (in
words and symbols), test statistic, p-value, and conclusion.

39. A person suspects that a certain grocery store is overstating the weight of fresh
chickens, and are thus overcharging their customers. He randomly samples fve
chicken packages, records the weight stated on the package (in grams) and the true
weight of the chicken. The results are illustrated in Table 10.10.

Even though it is somewhat dubious to use the t procedures for such a small sam-
ple size, the person conducting the investigation feels the normality assumption is
reasonable, and goes ahead with the t procedures. He fnd the output:
Balka ISE 1.09 10.7. CHAPTER EXERCISES 178

Chicken Stated weight Actual weight

1 908 866
2 1120 1087
3 795 783
4 912 890
5 1402 1397

Table 10.10: Stated weight and actual weight for a sample of fve chickens.

Paired t-test
data: stated and actual
t = 3.3861, df = 4, p-value = 0.02763
alternative hypothesis: true difference in means is not equal to 0
95 percent confidence interval:
4.104811 41.495189
sample estimates:
mean of the differences
22.8

(a) Give a summary of the results of the analysis.

(b) If the normality assumption is not in fact reasonable, what are the conse-
quences?

40. Researchers are investigating whether a new experimental method of measuring

viscosity in gases is as e˙ective as an older method. They take measurements of
viscosity on 6 samples of gas, using both the new method and the older method.
The results are given in Table 10.11. The researchers want to test whether there is
a di˙erence between the two methods, and estimate the size of the di˙erence with
a confdence interval.

Gas Sample Old Method New Method

1 2.74 2.71
2 3.11 3.01
3 2.99 3.11
4 2.99 2.94
5 2.88 2.96
6 3.29 3.21

Table 10.11: Measures of viscosity in gases.

(a) Explain why a paired-di˙erence procedure should be used in this situation.

(b) If we want to use a t inference procedure, how could we assess if the normality
assumption is reasonable?
For the following questions, suppose that the normality assumption is reason-
Balka ISE 1.09 10.7. CHAPTER EXERCISES 179

able.
(c) Calculate a 95% confdence interval for the population mean di˙erence, and
give a proper interpretation of the interval.
(d) Conduct the appropriate hypothesis test. Give appropriate hypotheses (in
words and symbols), test statistic, p-value, and conclusion.

41. The reaction times of 4 students are measured before and after having several
alcoholic beverages. The results are shown in Table 10.12. Suppose we wish to

Student Before After

1 0.41 0.78
2 0.38 0.63
3 0.52 0.94
4 0.41 0.47

Table 10.12: Before and after reaction times.

test the null hypothesis that alcohol has no e˙ect on reaction times, against the
alternative that it slows reaction times. Assume that the di˙erences in the times
are normally distributed, and use the appropriate t procedure. (But keep in mind
that the use of a t procedure is questionable for such a small sample size.)

(a) Carry out the hypothesis test. Give appropriate hypotheses, p-value, test
statistic and conclusion at the 5% signifcance level.
(b) Calculate a 95% confdence interval for the mean change in reaction times.
Give a proper interpretation of the interval.
(c) What are the advantages to setting up an experiment this way, instead of
splitting up the students into 2 independent groups?
(d) What are the disadvantages to setting up an experiment in this way?
Balka ISE 1.09 10.7. CHAPTER EXERCISES 180
Chapter 11

Inference for Proportions

J.B.’s strongly suggested exercises: 2, 5, 6, 8, 10, 12, 16, 17, 19, 20, 24, 25, 26, 27

11.1 Introduction

11.2 The Sampling Distribution of p̂

1. What are the mean and variance of the sampling distribution of p̂?

2. For what values of n and p does the normal approximation to the distribution of p̂
work best? For what values of n and p is the normal approximation very poor?

11.3 Confdence Intervals and Hypothesis Tests for the

Population Proportion p

3. When do we create a confdence interval for p̂?

4. A random sample of 200 observations from a population revealed that 62 individuals

had a certain characteristic.

(a) What is p̂?

(b) What is SE(p̂)?
(c) What is a 95% confdence interval for p?
(d) If we wish to test the null hypothesis that p = 0.5, what is SE0 (p̂)?

181
Balka ISE 1.09 11.4. DETERMINING THE MINIMUM SAMPLE SIZE N 182

(e) What is the value of the appropriate test statistic?

5. In each of the following scenarios, state if it is reasonable to use the normal approx-
imation to calculate a confdence interval for p. (Use the np̂ ≥ 15, n(1 − p̂) ≥ 15
guideline.)

(a) p̂ = 0.70, n = 10.

(b) p̂ = 0.70, n = 200.
(c) p̂ = 0.99, n = 200.
(d) p̂ = 0.01, n = 10,000.
(e) p̂ = 0, n = 200.

6. A study1 of births in Liverpool, UK, investigated a possible relationship between

parental smoking status during pregnancy and the likelihood of a male birth. In
one part of the study, researchers drew a sample of 363 births in which both parents
were heavy smokers during the pregnancy. Of the 363 babies born to these couples,
158 were male. Suppose we wish to construct a 95% confdence interval for the
population proportion of male births to heavy-smoking parents in Liverpool.

(a) What does p represent in this case?

(b) What is the point estimate of p?
(c) Is it reasonable to use large sample methods to calculate a confdence interval
for p here?
(d) What is the standard error of the sample proportion?
(e) What is a 95% confdence interval for p?
(f) Give an interpretation of the 95% confdence interval for p.
(g) Test the null hypothesis that the population proportion of male births to
heavy-smoking parents is 0.50, against a two sided alternative. Give appropri-
ate hypotheses, standard error, value of the test statistic, p-value, and conclu-
sion at α = 0.05.
(h) To what population do your conclusions apply?

11.4 Determining the Minimum Sample Size n

7. Find the minimum sample size required in each of the following situations.

(a) We wish to estimate p within 0.03 with 95% confdence, and we have no rea-
sonable estimate of p beforehand.
1
Koshy et al. (2010). Parental smoking and increased likelihood of female births. Annals of Human
Biology, 37(6):789–800.
 11.5.
Balka ISE 1.09 INFERENCE PROCEDURES FOR THE DIFFERENCE BETWEEN TWO POPULATION PROPORTIONS 183

(b) We wish to estimate p within 0.03 with 95% confdence, and from prior infor-
mation we feel strongly that p is approximately 0.20.
(c) We wish to estimate p within 0.01 with 90% confdence, and we have no rea-
sonable estimate of p beforehand.
(d) We wish to estimate p within 0.01 with 99% confdence, and we have no rea-
sonable estimate of p beforehand.
(e) We wish to estimate p within 0.01 with 99% confdence, and we know for
certain that p lies between 0.1 and 0.2.

11.5 Inference Procedures for the Di˙erence Between

Two Population Proportions

11.5.1 The Sampling Distribution of p̂1 − p̂2

11.5.2 Confdence Intervals and Hypothesis Tests for p1 − p2

8. In words, what is the meaning of SE(p̂1 − p̂2 )?

9. A random sample of 400 observations from population 1 revealed that 82 individuals

had a certain characteristic. A random sample of 400 observations from population
2 revealed that 104 individuals had that characteristic.

(a) What are the values of p̂1 and p̂2 ?

(b) What is SE(p̂1 − p̂2 )?
(c) Calculate a 95% confdence interval for p1 − p2 .
(d) Test the null hypothesis that the population proportions are equal, against a
two-sided alternative hypothesis. Give appropriate hypotheses, value of the
pooled proportion p̂, value of the standard error, test statistic and p-value. Is
there signifcant evidence against the null hypothesis at α = 0.05?

10. Much research has gone into studying how homing pigeons are able to navigate to
their home loft from an unfamiliar release point, but the precise mechanisms are
still unknown. It is known that homing pigeons can detect the earth’s magnetic
feld, and that is likely a contributing factor in their ability to navigate. A study2
investigated this in an experiment involving 77 homing pigeons. The pigeons were
randomly divided into a magnetic pulse group (group M) and a control group (group
C). The 38 control group pigeons were released at a location 106 km from the home
loft, and 22 found their way home. The 39 members of the magnetic pulse group
2
Holland et al. (2013). A magnetic pulse does not a˙ect homing pigeon navigation: a gps tracking
experiment. The Journal of Experimental Biology, 216:2192–2200.
Balka ISE 1.09 11.6. CHAPTER EXERCISES 184

received a strong magnetic pulse (perpendicular to the earth’s magnetic feld) before
being released from the same location. Twenty-one of the 39 magnetic pulse group
pigeons made it back to the home loft.

(a) Calculate a 95% confdence interval for the di˙erence between the population
proportion of pigeons that navigate home after being subjected to the magnetic
pulse, and the population proportion for the control group. (Hint to ease the
calculation burden: SE(p̂M − p̂C ) = 0.1131.)
(b) Give an appropriate interpretation of the interval found in 10a.
(c) Perform a hypothesis test of the null hypothesis that the two groups have
the same likelihood of making it back to their home loft. Give appropriate
hypotheses, value of the test statistic, p-value, and conclusion. (One could
make an argument for the one-sided alternative hypothesis that the magnetic
pulse reduces the probability of a pigeon arriving home, but play it safe and
use a two-sided alternative hypothesis.) (Hint to ease the calculation burden:
SE0 (p̂M − p̂C ) = 0.1132.)

11.6 Chapter Exercises

11.6.1 Basic Calculations

11. Calculate the p-value in the following situations.

(a) H0 : p = 0.3, Ha : p > 0.3, Z = −1.40.

(b) H0 : p = 0.3, Ha : p < 0.3, Z = −1.40.
(c) H0 : p = 0.3, Ha : p 6= 0.3, Z = −1.40.
(d) H0 : p = 0.6, Ha : p < 0.6, Z = −1.88.
(e) H0 : p = 0.6, Ha : p 6= 0.6, Z = −1.88.

11.6.2 Concepts

12. What is the di˙erence in meaning of the symbols p̂ and p?

13. Would it ever make sense to test H0 : p̂ = 0.25?

14. What is the meaning of the term SE(p̂)? Does the term SE(p) have a similar
meaning?

15. In each of the following scenarios, state whether using the normal approximation
to carry out a hypothesis test is reasonable. (Use the np0 ≥ 15, n(1 − p0 ) ≥ 15
Balka ISE 1.09 11.6. CHAPTER EXERCISES 185

guideline.)

(a) H0 : p = 0.20, n = 10.

(b) H0 : p = 0.20, n = 200.
(c) H0 : p = 0.9980, n = 10.
(d) H0 : p = 0.9980, n = 1,000.
(e) H0 : p = 0.9980, n = 100,000.

16. An Ipsos-Reid poll asked Canadians whether schools in their community have be-
come “less safe” than they were fve years ago. Fifty percent of the respondents said
schools were less safe than 5 years ago. The corresponding confdence interval for
p was found to be (0.46, 0.54).

There are some potential sources of bias in surveys like this one. For example, the
question could be biased (e.g. “With the increase in violent behaviour, schools have
become less safe, wouldn’t you agree?”) The folks at Ipsos are professionals, and
wouldn’t give questions this blatantly biased. But if our hope is to estimate the
proportion of all Canadians who feel that schools have becomes less safe, what are
some other potential sources of bias?

17. Test your conceptual understanding: Which of the following statements are true?
You should be able to explain why a statement is true or why a statement is false.

(a) p̂ is an unbiased estimator of p.

(b) When n < 30 we should use the t distribution when calculating confdence
intervals for p.
(c) The true distribution of p̂ is based on the binomial distribution.
(d) The true standard deviation of the sampling distribution of p̂ depends on the
value of p.
(e) The sampling distribution of p̂ is perfectly normal for large sample sizes.

18. Test your conceptual understanding: Which of the following statements are true?
You should be able to explain why a statement is true or why a statement is false.

(a) The normal approximation to the sampling distribution of p̂ works best when
we have a large sample size and p = 0.5.
(b) The sampling distribution of p̂ becomes more normal as p tends to 1.
(c) The sampling distribution of p is approximately normal for large sample sizes.
(d) All else being equal, the value of SE(p̂) decreases as the sample size increases.
(e) In repeated sampling, exactly 95% of 95% confdence intervals for p will capture
p.
Balka ISE 1.09 11.6. CHAPTER EXERCISES 186

11.6.3 Applications

19. Consider again the information in Question 10. In part of the experiment, 38 control
group pigeons were released from an unfamiliar site 106 km from their home loft,
and 22 of these pigeons were able to successfully navigate home. Suppose we wish
to construct a confdence interval for the true proportion of control group pigeons
that will fnd their way to their home loft (under the conditions of the study).

(a) What does p represent in this case?

(b) What is the point estimate of p?
(c) Is it reasonable to use large sample methods to calculate a confdence interval
for p here?
(d) What is the standard error of the sample proportion?
(e) What is a 95% confdence interval for p?
(f) Give a proper interpretation of the 95% confdence interval.
(g) Suppose that, before the study was conducted, a scientist claimed that the
true proportion of control group pigeons that would navigate back to their
home loft would be no more than 0.25. Does this study yield strong evidence
against this scientist’s claim? Test the null hypothesis that the population
proportion is 0.25, against the appropriate alternative hypothesis. Give ap-
propriate hypotheses, standard error, value of the test statistic, p-value, and
conclusion.

20. Seat belt marks (bruising) on the body are sometimes used as evidence that a person
was wearing a seat belt during a car crash. A study3 investigated seat belt marks
in victims of fatal car crashes in Sydney, Australia. (The authors looked only at
cases where there was no airbag deployment.) In 74 fatalities in which the victim
was wearing a seat belt, 27 victims showed seat belt marks.

(a) What does p represent in this case?

(b) What is the point estimate of p?
(c) Is it reasonable to use large sample methods to calculate a confdence interval
for p here?
(d) What is the standard error of the sample proportion?
(e) What is a 95% confdence interval for p?
(f) Give an appropriate interpretation of the confdence interval for p.
(g) Is there a hypothesis that should be tested here?

21. A study4 investigated various characteristics of Finnish murders. In one part of the
3
Chase et al. (2007). Safety restraint injuries in fatal motor vehicle collisions. Forensic Science, Medicine,
and Pathology, 3:258–263.
4
Häkkänen et al. (2009). Gender di˙erences in Finnish homicide o˙ence characteristics. Forensic Science
International, 186:75–80
Balka ISE 1.09 11.6. CHAPTER EXERCISES 187

study, a random sample of 91 Finnish male convicted murderers was drawn (there
was a single murder in each case). Twenty-four of these 91 murders were of the
o˙ender’s domestic partner. Suppose we wish to estimate the true proportion of
murders that are of the domestic partner of the o˙ender (for murders committed
by males).

(a) What does p represent in this case?

(b) What is the point estimate of p?
(c) Is it reasonable to use large sample methods to calculate a confdence interval
for p here?
(d) What is the standard error of the sample proportion?
(e) What is a 95% confdence interval for p?
(f) Give an interpretation of the 95% confdence interval for p.
(g) Suppose it was previously believed that in 40% of murders committed by
a Finnish male, the victim was the o˙ender’s domestic partner. Test the
null hypothesis that the population proportion is 0.40, against a two sided
alternative. Give appropriate hypotheses, standard error, value of the test
statistic, p-value, and conclusion at α = 0.05.
(h) To what population do the conclusions apply?

22. In another aspect of the study of Finnish murders frst discussed in Question 21,
researchers drew random samples of murders committed by males and murders
committed by females. Of 91 murders committed by a female, in 32 cases the
victim was their domestic partner. Of 91 murders committed by a male, in 24 cases
the victim was their domestic partner.

(a) Calculate a 90% confdence interval for the di˙erence in the population pro-
portions between female and male murderers. (Hint to ease the calculation
burden: SE(p̂F − p̂M ) = 0.0681.)
(b) Give an appropriate interpretation of the interval found in 22a.
(c) Perform a hypothesis test of the null hypothesis that, in Finnish murders, the
proportion of victims that are the domestic partner of the o˙ender is the same
for male and female o˙enders, against the alternative that these proportions
are di˙erent. Give appropriate hypotheses, value of the test statistic, p-value,
and conclusion at the 5% level of signifcance. (Hint to ease the calculation
burden: SE0 (p̂F − p̂M ) = 0.0684.)

23. A study5 at a medical clinic in Iran investigated a possible association between pep-
tic ulcers and a variety of factors. The study involved a sample of 60 peptic ulcer
su˙erers from a medical clinic, and a sample of 44 apparently healthy volunteers
to serve as the control group. In one part of the study, the authors investigated
5
Mahdavi et al. (2013). Is there any relationship between leb antigen expression and Helicobacter pylori
infection? Blood Cells, Molecules and Diseases, 51:174–176.
Balka ISE 1.09 11.6. CHAPTER EXERCISES 188

a possible di˙erence between the groups in their Lewis B (Leb ) blood group anti-
gen expression. (Previous studies had shown a possible relationship between Leb
expression and a mechanism for peptic ulcers.) Forty-three of the 60 peptic ulcer
su˙erers had Leb expression. Twenty-seven of the 44 control group members had
Leb expression.

(a) Calculate a 95% confdence interval for the di˙erence in the population pro-
portions between peptic ulcer patients and the control group. (Hint to ease
the calculation burden: SE(p̂U − p̂C ) = 0.0937.)
(b) Give an appropriate interpretation of the interval found in 23a.
(c) Perform a hypothesis test of the null hypothesis that the proportion of in-
dividuals with Leb expression is the same for the two groups, against the
alternative hypothesis that the proportions are di˙erent. Give appropriate hy-
potheses, value of the test statistic, p-value, and conclusion at the 5% level of
signifcance. (Hint to ease the calculation burden: SE0 (p̂U − p̂C ) = 0.0931.)

24. Consider again the study frst discussed in Question 6, which investigated a possible
relationship between parental smoking status during pregnancy and the likelihood
of a male birth. Of 363 male births in which both parents were heavy smokers
during the pregnancy, 158 of the babies were male. Of the 5045 births in which
both parents were non-smokers during the pregnancy, 2685 of the babies were male.

(a) Calculate a 95% confdence interval for the di˙erence between the population
proportion of male births for heavy-smoking parents, and the population pro-
portion of male births for non-smoking parents. (Hint to ease the calculation
burden: SE(p̂H − p̂N ) = 0.0270.)
(b) Give an appropriate interpretation of the interval found in 24a.
(c) Perform a hypothesis test of the null hypothesis that the male birth rate is
the same for both heavy-smoking and non-smoking parents. Give appropriate
hypotheses, value of the test statistic, p-value, and conclusion at the 5% level
of signifcance. (Hint to ease the calculation burden: SE0 (p̂H −p̂N ) = 0.0271.)
(d) To what population do your conclusions apply? Do the results of this study
imply strong evidence that parents smoking cause a decrease in the proportion
of male births?

25. Does a pleasant ambient fragrance increase the success rate of a courtship request?
A study6 investigated this question by having a good-looking 20 year-old man ap-
proach 18-25 year-old women walking alone in a mall. On 200 occasions, the man
approached as the woman was walking in an area of “pleasant ambient odours (e.g.,
pastries)”, and on 200 occasions he approached in an area of no odour. In every
case, the man approached the woman, made a brief scripted statement including a
6
Guéguen et al. (2012). The sweet smell of. courtship: E˙ects of pleasant ambient fragrance on womenÕs
receptivity to a manÕs courtship request. Journal of Environmental Psychology, 32:123–125
Balka ISE 1.09 11.6. CHAPTER EXERCISES 189

compliment, and then asked for the woman’s phone number. Of the 200 cases in
areas of pleasant ambient odour, 46 women gave their phone number. Of the 200
cases in areas of no ambient odour, 27 women gave their phone number.

(a) Calculate a 95% confdence interval for the di˙erence between the success rate
in areas with a pleasant fragrance and the success rate in areas with no odour
(pF − pC ) . (Hint to ease the calculation burden: SE(p̂F − p̂C ) = 0.0383.)
(b) Give an appropriate interpretation of the interval found in 25a.
(c) Perform a hypothesis test of the null hypothesis that the likelihood of getting a
phone number is the same in both areas. Give appropriate hypotheses, value
of the test statistic, p-value, and conclusion at the 5% level of signifcance.
(Hint to ease the calculation burden: SE0 (p̂F − p̂C ) = 0.0386.)

26. Part of the Salk polio vaccine trials in 1954 involved a massive double blind ran-
domized experiment, in which approximately 400,000 elementary school children
were randomly assigned to either the Salk polio vaccine or a placebo group. The
results:
Placebo Vaccine
Polio 160 86
No polio 200270 199661
Total 200430 199747

(The counts have been slightly modifed from the original data.)

(a) Let pp represent the population proportion of those in the placebo group that
would develop polio under these conditions, and pv represent the population
proportion of those in the vaccine group that would develop polio under these
conditions. What is the point estimate of the di˙erence in the population
proportions? (What is the point estimate of pp − pv ?)

(b) If we wish to test the null hypothesis that the population proportions are
equal, what is the value of the appropriate Z test statistic? (Hint to ease the
calculation burden: SE0 (p̂p − p̂v ) = 0.00007836327.)

(c) What is the p-value of the test of the null hypothesis of equal population
proportions against a two-sided alternative hypothesis?

(d) Give a summary of the results of the hypothesis test.

(e) Since polio is a rare disease, the di˙erence in population proportions may
not be the best way of comparing the two proportions. In many situations,
especially with rare events, the relative risk is the more informative quantity.
In this case the relative risk would be the ratio of the estimated probabilities
p̂placebo
of developing polio. That is, RR = p̂vaccine . If we have a relative risk of 3, say,
Balka ISE 1.09 11.6. CHAPTER EXERCISES 190

then that means individuals given the placebo would be 3 times more likely
to develop polio. In many cases this is a better and more intuitive measure
of the di˙erence between proportions. What is the relative risk of developing
polio?

11.6.4 Extra Practice Questions

27. The median lethal dose (often called LD50 (lethal dose, 50%) or LC50 (lethal con-
centration, 50%)), is often reported in toxicology studies. It is the dose that results
in 50% mortality within a specifed time period. Suppose you are investigating the
e˙ect of zinc concentration on the mortality of a certain type of freshwater minnow.
As part of this study, you subject 100 of these minnows to a zinc concentration of
5 ppm. Within 48 hours, 68 of the minnows are dead.

(a) Construct a 95% confdence interval for the true proportion of this type of
minnow that would be dead within 48 hours if exposed to 5 ppm of zinc.

(b) Based on this interval, is there a strong indication that the 48 hour LD50 is
di˙erent from 5 ppm?

(c) Carry out a hypothesis test of the null hypothesis that 5 ppm of zinc will kill
50% of this type of minnow within 48 hours, against a two-sided alternative
hypothesis. Give appropriate hypotheses, standard error, test statistic, p-
value, and conclusion at α = 0.05.

(d) Suppose in a di˙erent experiment, with a di˙erent type of minnow, we wished

to estimate the proportion that would be dead in 48 hours to within 0.03, with
99% confdence. What is the minimum sample size that would be required?
(Suppose that we have no reasonable estimate of the true proportion before
the study, and thus we must use the conservative estimate of p in the formula.)

(e) Suppose that in another part of this study, after exposure to a high level of
zinc, 198 of 200 minnows are dead within 48 hours. Would it be reasonable to
use the normal approximation to construct a confdence interval for p in this
scenario?

28. Suppose it was previously believed that in a certain geographical area, approxi-
mately 50% of babies born prematurely at 25 weeks did not survive until at least
their frst birthday. You investigate this claim and fnd data on 67 babies born at
25 weeks. 43 of these babies survived until they were one year old. Suppose it is
reasonable to assume that your data represents a random sample of births from
that area at that time.
Balka ISE 1.09 11.6. CHAPTER EXERCISES 191

(a) Construct a 95% confdence interval for the proportion of babies born at 25
weeks that survive until they are at least one year old. Give an appropriate
interpretation of the interval.
(b) Test the null hypothesis that half of babies born at 25 weeks die, against a
two-sided alternative hypothesis. Give appropriate hypotheses, test statistic,
p-value, and conclusion.

29. A random sample of 40 soft-serve ice cream vendors in a large city revealed that 17
of them had E. coli counts in excess of recommended guidelines. Calculate a 95%
confdence interval for the population proportion, and give a proper interpretation
of the interval.

30. The 2006 Canadian census revealed that approximately 23% of Canadians between
25 and 64 years of age have a university degree. In a certain area at that time, a
random sample of 500 adults in this age group revealed that 25 had a university
degree.

(a) Construct a 90% confdence interval for the proportion of people between 25
and 64 in this area at that time that had a university degree.
(b) Test whether the proportion of people between 25 and 64 in this area with a
university degree di˙ered from the rest of Canada. Give appropriate hypothe-
ses, test statistic, p-value, and conclusion.

31. You are interested in using a marketing campaign to sell new cars to recent uni-
versity graduates. Before starting the campaign, you would like to have some idea
of the proportion of graduating students who intend to buy a new car in the next
year. If you wish to estimate this proportion to within 0.04 with 90% confdence,
what is the minimum sample size that is required? Assume that you do not have
a good estimate of the proportion before taking the sample, and thus you need to
use the most conservative estimate of the sample size required.

32. Pollsters were interested in possible di˙erences between men and women in their
presidential approval ratings. 100 men and 100 women were asked if they approved
of the way the president was handling his job. 58 of the women and 43 of the men
said they approved.

(a) Calculate a 90% confdence interval for the di˙erence in the population pro-
portions between females and males. (Hint to ease the calculation burden:
SE(p̂F − p̂M ) = 0.0699.)
(b) Give an appropriate interpretation of the interval found in 32a.
(c) Perform a hypothesis test of the null hypothesis that the approval ratings for
men and women are the same, against the alternative that they are di˙erent.
Give appropriate hypotheses, value of the test statistic, p-value, and a con-
clusion at the 5% level of signifcance. (Hint to ease the calculation burden:
Balka ISE 1.09 11.6. CHAPTER EXERCISES 192

SE0 (p̂F − p̂M ) = 0.070707.)

33. A company is testing a new heat treatment intended to eradicate bedbugs. As part
of the investigation process, they wish to assess what e˙ect di˙erent temperatures
have on killing bedbugs. They subject 200 adult bedbugs to a temperature of 45
degrees Celsius for 30 minutes, and fnd that 176 died. They subject 200 di˙erent
adult bedbugs to a temperature of 47 degrees Celsius for 30 minutes, and fnd that
184 died.

(a) What is a 95% confdence interval for the di˙erence in population proportions?
Also give an appropriate interpretation of the interval. (Hint to ease the
calculation burden: SE(p̂1 − p̂2 ) = 0.02993.)
(b) Suppose we wish to test the null hypothesis that the population proportions
are equal, against the alternative that the higher temperature kills a higher
proportion of bedbugs. Give appropriate hypotheses in words and symbols,
standard error, value of the test statistic, and conclusion at a α = 0.01 signif-
icance level. (Hint to ease the calculation burden: SE0 (p̂1 − p̂2 ) = 0.030.)
Chapter 12

Inference for Variances

(No exercises yet.)

193
Balka ISE 1.09 194
Chapter 13

χ2 Tests for Count Data

J.B.’s strongly suggested exercises: 2, 3, 4, 6, 7, 8, 10, 11, 14, 15, 16, 17, 19, 21, 22, 23

13.1 Introduction

13.2 χ2 Tests for One-Way Tables

13.2.1 The χ2 Test Statistic

1. Suppose we have a sample of 400 observations that are categorized into 4 groups.
The frst group contains 52 of the observations, the second 149, the third 99, and
the fourth 100 observations. Suppose that we wish to test the null hypothesis that
these four groups are equally likely.

(a) What are the appropriate hypotheses, in both words and symbols?
(b) What are the observed counts? What are the expected counts under the null
hypothesis?
(c) What is the value of the test statistic?
(d) What are the appropriate degrees of freedom?
(e) What is the p-value of the test?
(f) Can the null hypothesis be rejected at α = 0.10? At α = 0.01?

2. Does the distribution of birth dates of Canadian NHL players di˙er from that of
Canadians as a whole? In his popular book Outliers, Malcolm Gladwell made the
claim that Canadian children born earlier in the year (January, February, March)

195
Balka ISE 1.09 13.2. χ2 TESTS FOR ONE-WAY TABLES 196

have a better chance of making the NHL when compared to children born later in the
year. Does NHL data from 2010/2011 support this claim? Table 13.1 illustrates the
distribution of a sample of 510 Canadian NHL player birthdays and the distribution
for all Canadians.12

Birth date Q1 Q2 Q3 Q4 Total

Count (NHL) 135 146 124 105 510
Percentage (NHL) 26.5 28.6 24.3 20.6 100
Percentage (Canadians) 24.2 26.0 26.0 23.8 100

Table 13.1: Birthdays of Canadian NHL players and Canadians by quarter.

(a) Is the observed percentage of Q1 birthdays higher for Canadian NHL players
than for Canadians as a whole? Is the observed percentage of Q4 birthdays
lower for Canadian NHL players than for Canadians as a whole?
(b) Suppose we wish use a χ2 test to test the hypothesis that the distribution of
birthdays for Canadian NHL players is the same as that of Canadians as a
whole. What are the appropriate hypotheses of the test?
(c) If H0 were true, what would be the expected count of NHL player birthdays
in Q1? Is this higher or lower than the observed count?
(d) What are the degrees of freedom for the χ2 test statistic?
(e) Calculate the value of the χ2 test statistic.
(f) What is the p-value of the test?
(g) Give an appropriate conclusion to the hypothesis test at the 5% signifcance
level.

13.2.2 Testing Goodness-of-Fit for Specifc Parametric Distribu-

tions

3. Suppose that a type of bird lays exactly three eggs per year (this is unrealistic of
course, but play along). A scientist claims that the number of eggs that hatch per
nest follows a binomial distribution with n = 3, p = 0.80. You do not believe this
scientist, and follow a sample of 500 nests until the eggs hatch. The results for the
500 nests are summarized in the Table 13.2.

Does this sample provide strong evidence against the scientist’s claim that the
number of eggs that hatch in a nest has a binomial distribution with n = 3, p =
1
The NHL birth date data was pulled on March 24 2011 using a search on www.nhl.com for current NHL
players with Canadian birthplaces.
2
Q1 is defned to be January, February March, Q2 is defned to be April, May, June, Q3 is defned to be
July, August, September, Q4 is defned to be October, November, December. The distribution of birth
dates for all Canadians was based on information from Statistics Canada.
Balka ISE 1.09 13.3. χ2 TESTS FOR TWO-WAY TABLES 197

Number that hatch 0 1 2 3

Frequency 80 29 164 227

Table 13.2: Number of eggs that hatch per nest.

0.80? Test the claim by performing a chi-square goodness-of-ft test. (Hint: To get
the expected counts under the null hypothesis, use the binomial formula with the
hypothesized values (n = 3, p = 0.80), and multiply the probabilities by 500. Note
that sometimes we use the data to estimate p, but here the value of p is specifed
in the null hypothesis.)

(a) What are the appropriate hypotheses?

(b) What are the expected counts?
(c) What is the value of the appropriate test statistic?
(d) What are the appropriate degrees of freedom?
(e) What is the p-value of the test?
(f) Give an appropriate conclusion at the 5% signifcance level.

4. Consider again the information in the Question 3. Suppose a di˙erent scientist has
claimed that the number of eggs that hatch should follow a binomial distribution
(with an unknown value of p). Does the sample data given in the previous question
provide strong evidence against the scientist’s claim that the number of eggs that
hatch per nest follows a binomial distribution? Test the claim by performing a χ2
goodness-of-ft test. (The di˙erence between this question and the previous one is
that here we must use the data to estimate p, which is the more likely scenario in
the real world.)

5. Consider again the information in the Question 3. Suppose a di˙erent scientist has
suggested that the 4 possible outcomes (0 hatch, 1 hatch, 2 hatch, 3 hatch) are all
equally likely. Using the data give in Question 3, test the null hypothesis that the 4
possible outcomes are all equally likely. Give appropriate hypotheses, test statistic,
p-value, and conclusion.

13.3 χ2 Tests for Two-Way Tables

6. Suppose we have a random sample of 1000 observations, categorized according to

two variables. The results are given in Table 13.3.

(a) What are the appropriate hypotheses?

(b) What are the expected counts in each cell?
(c) What is the value of the test statistic?
Balka ISE 1.09 13.3. χ2 TESTS FOR TWO-WAY TABLES 198

C1 C2
R1 50 150
R2 200 20
R3 400 180

Table 13.3: A 3 × 2 contingency table.

(d) What are the appropriate degrees of freedom?

(e) What is the p-value of the test?
(f) Is the null hypothesis rejected at α = 0.10? At α = 0.01?

7. Many cities in the United States have buyback programs for handguns, in which
the police department pays people to turn in guns. The guns are then destroyed.
Is there a di˙erence between the distribution of the size of guns turned in during
buyback programs and the distribution of the size of guns used in homicides and
suicides? A study3 investigated this question, using data from a gun buyback
program and police records in Milwaukee. Table 13.4 and Figure 13.1 illustrate the
distribution of the calibre of the gun (small, medium, large, other) in the di˙erent
scenarios.

Gun Calibre Buybacks Homicides Suicides Total

Small 719 75 40 834
Medium 182 202 72 456
Large 20 40 13 73
Other 20 52 0 72
Total 941 369 125 1435

Table 13.4: Gun calibre for buyback guns and guns used in homicides and suicides.

1.0
0.7
Other Other
Relative Frequency

Relative Frequency

0.6 0.8
Large Large
0.5 Medium Medium
Small 0.6 Small
0.4

0.3 0.4

0.2
0.2
0.1

0.0 0.0
Buybacks Homicides Suicides Buybacks Homicides Suicides

(a) Side-by-side barcharts. (b) Stacked barcharts.

Figure 13.1: The distribution of calibre of gun for buybacks, homicides, and suicides.
3
Kuhn, E., Nie, C., O’Brien, M., Withers, R. L., Wintemute, G., and Hargarten, S. W. (2002). Missing
the target: a comparison of buyback and fatality related guns. Injury Prevention, 8:143–146
Balka ISE 1.09 13.4. CHAPTER EXERCISES 199

(a) What percentage of buyback guns were of small calibre? What percentage of
homicide guns were of small calibre?
(b) Suppose we wish to investigate a possible association between gun calibre
and gun type (buyback, homicide, or suicide) using a χ2 test. What are the
appropriate hypotheses of the test?
(c) If H0 were true, what would be the expected count of small calibre buyback
guns?
(d) What are the degrees of freedom for the χ2 test statistic?
(e) The value of the χ2 test statistic is 422.48. (You should be able to calculate
this value, but the calculations are rather long and tedious.) What is the
p-value of the test?
(f) Give an appropriate conclusion to the hypothesis test.
(g) Looking at Table 13.4 and Figure 13.1, what are the major di˙erences between
the distribution of calibre of gun between the buyback guns and the guns used
in homicides or suicides?
(h) To what population do your conclusions apply?

13.4 Chapter Exercises

13.4.1 Basic Calculations

13.4.2 Concepts

8. What will happen if we carry out a χ2 test and one or more of our expected counts
is very small (less than one, say)?

(a) The world will end.

(b) We cannot calculate the test statitistic.
(c) The test statistic will be negative.
(d) Our conclusion will be the exact opposite of what it should be.
(e) The χ2 approximation may not be very good, possibly resulting in misleading
conclusions.

9. Fifty children were asked to taste 5 di˙erence ice creams, and were asked which
favours they liked. The results:
Vanilla Chocolate Cherry & Mango Double Mango “All Natural Mango”
(with simulated mango favour)
32 42 16 15 14

Would it be reasonable to use our usual χ2 test here? Why or why not?
Balka ISE 1.09 13.4. CHAPTER EXERCISES 200

10. (a) If we let X represent a random variable that has a χ2 distribution with 8
degrees of freedom, what is P (X > 21.955)?
(b) If we let X represent a random variable that has a χ2 distribution with 2
degrees of freedom, what is P (X > 3.794)?
(c) If Z1 , Z2 , Z3 , Z4 are independent standard normal random variables, what is
P (Z12 + Z22 + Z32 + Z42 > 7.779)?
(d) If Z1 , Z2 , Z3 are independent standard normal random variables, what is P (Z12 + Z22 + Z32 < 11.345)?

11. Test your conceptual understanding: Which of the following statements are true?
You should be able to explain why a statement is true or why a statement is false.

(a) In χ2 tests for count data, all else being equal, the greater the value of the test
statistic the smaller the p-value.
(b) The χ2 distribution tends toward the standard normal distribution as the
degrees of freedom increase.
(c) The χ2 distribution is skewed to the right.
(d) In χ2 tests for count data, we lose 1 degree of freedom for each parameter we
estimate using our data.
(e) χ2 statistics can take on negative values.

12. Test your conceptual understanding: Which of the following statements are true?
You should be able to explain why a statement is true or why a statement is false.

(a) The χ2 test statistic has exactly a χ2 distribution.

(b) The mean of a χ2 distribution is equal to its degrees of freedom.
(c) For a given set of expected counts in a two-way table of a given size, the
greater the di˙erence between the observed and expected counts, the greater
the value of the χ2 test statistic.
(d) For a given set of expected counts in a two-way table of a given size, the
greater the di˙erence between the observed and expected counts, the smaller
the p-value.
(e) For a given set of expected counts in a two-way table of a given size, the
greater the di˙erence between the observed and expected counts, the greater
the evidence against the null hypothesis.
(f) For a 2 × 2 table, the value of the χ2 test statistic is the square of the corre-
sponding Z test statistic.
Balka ISE 1.09 13.4. CHAPTER EXERCISES 201

13.4.3 Applications

13. In a study4 investigating genetic linkage in tomato plants, the researchers crossed
one pure line of tomato plants that were tall and cut-leaf with another pure line that
were dwarf and potato-leaf. Tall is genetically dominant to dwarf, and cut-leaf is ge-
netically dominant to potato-leaf, so the tomato plants resulting from this frst gen-
eration cross were all tall and cut-leaf (they had the tall/cut-leaf phenotype). This
frst generation was then self-crossed. For the second generation, under Mendelian
inheritance with independent assortment, a 9:3:3:1 ratio of tall/cut-leaf:tall/potato-
leaf:dwarf/cut-leaf:dwarf/potato-leaf phenotypes would be expected. A ratio very
di˙erent from this would indicate that the gene controlling the height and the gene
controlling the leaf type are not being inherited independently. The counts of ob-
served phenotypes for the 1611 tomato plants are given in Table 13.5.

Phenotype
Tall/cut-leaf Tall/potato-leaf Dwarf/cut-leaf Dwarf/potato-leaf
Observed count 926 288 293 104

Table 13.5: Phenotypes observed in the second generation.

Suppose we wish to test the null hypothesis that these 4 phenotypes occur in a
9:3:3:1 ratio.

(a) What are the appropriate hypotheses?

(b) What is the value of the χ2 test statistic?
(c) What is the p-value of the test?
(d) Give an appropriate conclusion to the test.

14. In genetics experiments involving pea plants (like the famous ones conducted by
Mendel), one variable that is often measured is the shape of the ripe seeds (the
seeds are either round or wrinkled). Round seeds (R) are dominant over wrinkled
seeds (r) so a pea plant will have round seeds if it is homozygous for round seeds
(RR) or heterozygous (Rr), and will have wrinkled seeds only if it is homozygous
for wrinkled seeds (rr). If heterozygous plants are self-crossed, then on average 75%
of the resulting plants will have round seeds and 25% will have wrinkled seeds.

Suppose you suspect that a feld of pea plants is heterozygous for seed shape (Rr),
but you are not sure. (Perhaps you were conducting an experiment and lost your
notes about what was on that feld.) These plants are self-crossed, and of the
resulting 80 plants, 72 have round seeds and 8 have wrinkled seeds. Does this
give strong evidence that the true proportion of plants with round seeds from this
crossing di˙ers from 0.75? You may wish to use a χ2 test to investigate this question.
4
MacArthur, J. (1931). Linkage studies with the tomato. Transactions of the Royal Canadian Institute,
18:1–19.
Balka ISE 1.09 13.4. CHAPTER EXERCISES 202

(a) What are the appropriate hypotheses?

(b) What is the value of the χ2 test statistic?
(c) What is the p-value of the test?
(d) What is an appropriate conclusion at the 5% signifcance level?

15. Consider again the information in Question 14. In this question the null hypothesis
that the true proportion of plants with round seeds is 0.75 was tested with a χ2
test. This null hypothesis could also be tested with a Z test for a single proportion.

(a) What is the value of the Z test statistic?

(b) What is the p-value of the Z test for a two-sided alternative hypothesis?
(c) What is the relationship between the Z statistic and the χ2 test statistic found
in 14b?
(d) What is the relationship between the p-value of the Z test and the p-value of
the χ2 test found in part 14c?

16. Benford’s law describes the distribution of the frst digit in the value of a variable.
(The frst digit refers to the frst non-zero digit in the number. For example, the
frst digit of 312.7 is 3, and the frst digit of 0.072 is 7.) In many naturally occurring
scenarios, the digits 1–9 are not equally likely to be the frst digit (smaller frst digit
values are more likely). This has been shown for many di˙erent variables, including
bank account balances, populations of cities around the world, and lengths of rivers.
(It does not apply to all types of data. For example, artifcially structured numbers
like telephone numbers and zip codes do not have this feature.) For variables that
satisfy Benford’s law, the probability that the frst digit is d (d = 1, . . . , 9) is
log10 (1 + d1 ). Under Benford’s law the frst digit has approximately a 30% chance
of being a 1, but only a 5% chance of being a 9.

What is the distribution of the frst digit in sample means reported in scientifc
journal articles? Does the distribution follow Benford’s law? Or is it closer to a dis-
crete uniform distribution, or something else entirely? A sample of 126 means from
journal articles resulted in the frst digits reported in Table 13.6 and Figure 13.2.
(The means were independently drawn by your author from a variety of scientifc
journal articles.)

(a) Suppose we wish to test the null hypothesis that the distribution of the frst
digit in reported sample means is uniform. What are the appropriate hypothe-
ses?
(b) What are the degrees of freedom?
(c) The value of the χ2 test statistic is 49.29. (The calculation is straightforward,
and you should understand how to arrive at this value, but it is a bit of a
hassle to calculate.) What is the p-value of the test?
(d) What is an appropriate conclusion to the hypothesis test?
Balka ISE 1.09 13.4. CHAPTER EXERCISES 203

First digit 1 2 3 4 5 6 7 8 9
Observed count 34 22 17 11 11 13 7 4 7
Observed proportion 0.27 0.17 0.13 0.09 0.09 0.10 0.06 0.03 0.06
Uniform proportion 1/9 1/9 1/9 1/9 1/9 1/9 1/9 1/9 1/9
Uniform expected count 14 14 14 14 14 14 14 14 14
Benford proportion 0.30 0.18 0.12 0.10 0.08 0.07 0.06 0.05 0.05
Benford expected count 37.9 22.2 15.7 12.2 10.0 8.4 7.3 6.4 5.8

Table 13.6: Observed counts and proportions, and the counts and proportions that would
be expected under a uniform distribution and under Benford’s law for 126 sample means.
0.30

Observed
Uniform
Benford
Proportion
0.20
0.10
0.00

1 2 3 4 5 6 7 8 9
First digit

Figure 13.2: The observed proportion, and the expected proportions under a uniform
distribution and Benford’s law.

(e) The hypothesis test yields extremely strong evidence that the true distribu-
tion is not uniform. Use information from Figure 13.2 to describe how the
distribution di˙ers from the uniform distribution.
(f) Suppose we wish to test the null hypothesis that the distribution of the frst
digit in reported sample means follows Benford’s law. What are the appropri-
ate hypotheses?
(g) What are the degrees of freedom?
(h) The value of the χ2 test statistic is 4.41. (The calculation is straightforward,
and you should understand how to arrive at this value, but it is a bit of a
hassle to calculate.) What is the p-value of the test?
(i) What is an appropriate conclusion?
(j) Are the results of the test consistent with what is illustrated in Figure 13.2?

17. A study5 involving a random sample of 305 Libyan medical students revealed the
following information about their ABO blood type and Rh factor.
5
Fayrouz et al. (2012). Relation between fngerprints and di˙erent blood groups. Journal of Forensic
and Legal Medicine, 19:18–21.
Balka ISE 1.09 13.4. CHAPTER EXERCISES 204

A B AB 0 Total
Rh +ve 92 30 8 136 266
Rh -ve 9 9 8 13 39
Total 101 39 16 149 305

Table 13.7: Blood types of 305 Libyan medical students.

(a) Suppose we wish to investigate a possible association between ABO blood type
and Rh factor using a χ2 test. What are the appropriate hypotheses of the
test?
(b) If H0 were true, what would be the expected count of individuals with blood
type A −ve?
(c) What are the degrees of freedom for the χ2 test statistic?
(d) The value of the χ2 test statistic is 27.1367. (You should be able to calculate
this value, but the calculations are rather long and tedious.) What is the
p-value of the test?
(e) Give an appropriate conclusion at the 5% signifcance level.
(f) To what population do your conclusions apply?
(g) Optional: Why is the conclusion to the test an unusual one? (This requires
some background knowledge in blood types.)

18. An Iranian study6 investigated a possible association between peptic ulcers and a
variety of factors. The study involved a sample of 60 peptic ulcer su˙erers from
a medical clinic, and a sample of 44 apparently healthy volunteers to serve as the
control group. In one part of the study, the ABO blood type of the individuals was
recorded. The results are listed in Table 13.8.

A B AB O
Peptic ulcer su˙erers 11 16 3 30
Healthy individuals 15 5 5 19

Table 13.8: Blood types of 60 patients with peptic ulcers and 44 healthy individuals.

(a) What percentage of the 60 of peptic ulcer patients have blood type A? What
percentage of the 44 healthy individuals have blood type A?
(b) Suppose we wish to investigate a possible association between peptic ulcers
and ABO blood type using a χ2 test. What are the appropriate hypotheses of
the test?
(c) If H0 were true, what would be the expected count of peptic ulcer patients
with blood type A?
6
Mahdavi et al. (2013). Is there any relationship between leb antigen expression and Helicobacter pylori
infection? Blood Cells, Molecules and Diseases, 51:174–176.
Balka ISE 1.09 13.4. CHAPTER EXERCISES 205

(d) What are the degrees of freedom for the χ2 test statistic?
(e) The value of the χ2 test statistic is 7.0521. (You should be able to calculate
this value, but the calculations are rather long and tedious.) What is the
p-value of the test?
(f) Give an appropriate conclusion to the hypothesis test at the 5% level of sig-
nifcance.

19. A study7 of births in Liverpool, UK, investigated a possible relationship between

parental smoking status during pregnancy and the likelihood of a male birth. The
results for 6853 births in which the father was a non-smoker are given in Table 13.9.

No smoking Light smoking Heavy smoking Total

Male birth 2685 735 175 3595
Female birth 2360 708 190 3258
Total 5045 1443 365 6853

Table 13.9: Maternal smoking status during pregnancy and sex of the child for 6853
births. (The father was a non-smoker in each case.)

(a) What is the estimated proportion of male births for non-smoking couples?
What is the estimated proportion of male births for couples in which the
father is a non-smoker, but the mother is a heavy smoker?
(b) Suppose we wish to investigate a possible association between maternal smok-
ing status and sex of the child using a χ2 test. What are the appropriate
hypotheses of the test?
(c) If H0 were true, what would be the expected count of male births to non-
smoking women?
(d) What are the degrees of freedom for the χ2 test statistic?
(e) The value of the χ2 test statistic is 5.499. (You should be able to calculate this
value, but the calculations are rather long and tedious.) What is the p-value
of the test?
(f) Give an appropriate conclusion to the hypothesis test.
(g) To what population do your conclusions apply?

20. In the same study as given in Question 19, the researchers also investigated a
possible association between paternal smoking status during pregnancy and the
likelihood of a male birth. The results for 6032 births in which the mother was a
non-smoker are given in Table 13.10.

(a) What is the estimated proportion of male births for non-smoking couples?
What is the estimated proportion of male births for couples in which the
7
Koshy et al. (2010). Parental smoking and increased likelihood of female births. Annals of Human
Biology, 37(6):789–800.
Balka ISE 1.09 13.4. CHAPTER EXERCISES 206

No smoking Light smoking Heavy smoking Total

Male birth 2685 385 122 3192
Female birth 2360 357 123 2840
Total 5045 742 245 6032

Table 13.10: Paternal smoking status during pregnancy and sex of the child for 6032
births. (The mother was a non-smoker in each case.)

mother is a non-smoker, but the father is a heavy smoker?

(b) Suppose we wish to investigate a possible association between paternal smok-
ing status and sex of the child using a χ2 test. What are the appropriate
hypotheses of the test?
(c) If H0 were true, what would be the expected count of male births to a non-
smoking father?
(d) What are the degrees of freedom for the χ2 test statistic?
(e) The value of the χ2 test statistic is 1.461. (You should be able to calculate this
value, but the calculations are rather long and tedious.) What is the p-value
of the test?
(f) Give an appropriate conclusion to the hypothesis test.
(g) To what population do your conclusions apply?

21. A Finnish study8 looked at a possible association between the gender of convicted
murderers and their relationship to the victim. A random selection of 91 female
murderers and a random selection of 91 male murderers revealed the information
in Table 13.11.

Gender Male Female Total

Acquaintance 61 37 98
Partner 22 32 54
Family Member 4 18 22
Stranger 4 4 8
Total 91 91 182

Table 13.11: Gender of murderer and their relationship to the o˙ender in Finnish murders.

(a) For female murderers, what percentage of murders were of family members?
For male murderers, what percentage of murders were of family members?
(b) Suppose we wish to investigate a possible association between the gender of
the murderer and the relationship to the victim using a χ2 test. What are the
appropriate hypotheses of the test?
8
Häkkänen et al. (2009). Gender di˙erences in Finnish homicide o˙ence characteristics. Forensic Science
International, 186:75–80.
Balka ISE 1.09 13.4. CHAPTER EXERCISES 207

(c) If H0 were true, what would be the expected count for female murderers of
family members?
(d) What are the degrees of freedom for the χ2 test statistic?
(e) The value of the χ2 test statistic is 16.639. (You should be able to calculate
this value, but the calculations are rather long and tedious.) What is the
p-value of the test?
(f) Give an appropriate conclusion to the hypothesis test.
(g) By looking at the values in the table, what appear to be the di˙erences in the
distributions of relationship to the victim between male and female murderers?
(h) To what population do your conclusions apply?

22. Consider again the information from the Polio vaccine trials (which was frst dis-
cussed in the exercises for inference for proportions).

Placebo Vaccine
Polio 160 86
No polio 200270 199661
Total 200430 199747

(a) Carry out a χ2 test, testing the null hypothesis that the vaccine has no e˙ect.
What is the value of the test statistic?
(b) When we analyzed this data using a two sample Z test, we found Z = 4.6927.
What is the relationship between the two-sample Z test statistic value the χ2
test statistic value found here?

23. In order to make it a little more diÿcult for test writers to cheat, a professor has
two versions of each test. One version has a white cover sheet and one has a yellow
cover sheet. The versions are of equal diÿculty level, but the questions are slightly
di˙erent and the order is changed. In the examination room, students choose where
to sit and thus choose whether to write the white or yellow version (the tests were
placed around the rooms beforehand).

If a student’s ID number is not flled out properly on the computer answer sheet,
then it must be tracked down and fxed. In one test, far more writers of the yellow
exam gave an incorrect ID number, as illustrated in Table 13.12.

White Yellow
Incorrect ID 1 16
Correct ID 422 403

Table 13.12: Number of correctly and incorrectly flled out ID numbers on a statistics
test. (This is real data from a 2012 course at the University of Guelph.)

Suppose we wish to test whether there is an association between the colour of the
Balka ISE 1.09 13.4. CHAPTER EXERCISES 208

test and whether or not the student correctly flled in their ID number.

(a) What are the appropriate hypotheses?

(b) What is the value of the χ2 test statistic?
(c) What is the p-value of the test?
(d) Even though the test yields very strong evidence against the null hypothesis,
the professor believed it to simply be a fuke and not a real e˙ect. What are
some factors that might make them feel this way?

13.4.4 Extra Practice Questions

24. One hundred volunteers take part in a psychology study. The participants are shown
3 photos of an individual, one after the other, and asked which picture is the most
fattering. In reality, all 3 pictures are the same. In the study, 24 individuals picked
picture #1, 37 picked picture #2, and 39 picked picture #3. Test the hypothesis
that all 3 pictures are equally preferred.

(a) What are the appropriate hypotheses, in words and symbols?

(b) What is the value of the test statistic?
(c) What is the p-value of the test?
(d) Give an appropriate conclusion.

25. Viewers of a newly released movie were asked to give it a rating from 1 to 4 stars.
The results from a sample of 400 viewers are illustrated in Table 13.13.

# of stars 1? 2? 3? 4?
# of viewers 12 79 209 100

Table 13.13: Ratings of 400 viewers.

Before the sample was drawn, a movie executive predicted that viewers would give
the above ratings in a 1:1:2:2 ratio. That is, she believed the proportions would
be 1/6, 1/6, 2/6, 2/6. Does the sample above provide strong evidence against the
executive’s claim? Perform the appropriate chi-square test.

(a) What are the appropriate hypotheses, in words and symbols?

(b) What is the value of the test statistic?
(c) What is the p-value?
(d) Give an appropriate conclusion.

26. A manufacturer of potato chips wants to test the popularity of 3 new favours it
has developed. As an initial test, the product development branch wants to test
the null hypothesis that all 3 brands are equally preferred by consumers. They
Balka ISE 1.09 13.4. CHAPTER EXERCISES 209

administered a taste test to 120 people, in which each person tasted all 3 potato
chip favours and named their favourite. The results are given in Table 13.14.

Flavour Salty Tobacco Sweet BBQ Raspberry Vinaigrette

# who preferred favour 26 67 27

Table 13.14: Favourite chip favours in a taste test.

Suppose we wish to test the null hypothesis that all 3 favours are equally pre-
ferred. Carry out the hypothesis test. Give appropriate hypotheses, value of the
test statistic, p-value, and conclusion at a 5% signifcance level.

27. For a newly released movie, executives wanted to investigate a possible relationship
between a viewer’s age and their rating of a movie. Table 13.15 illustrates the
results.

Age / Rating 1? 2? 3? 4?
15–30 23 66 112 317
31–45 37 42 87 23
≥ 46 87 102 37 21

Table 13.15: Ratings vs age for movie goers.

The executives wanted to test the null hypothesis that age and rating are indepen-
dent, and they carried out an appropriate χ2 test. The value of the appropriate χ2
test statistic was found to be 377.42. (You should know how to calculate this value,
but it would be very time consuming to actually calculate it.) Find the appropriate
p-value and give an appropriate conclusion.

28. Researchers are investigating the properties of four types of concrete. The four
types are subjected to water penetration and cycles of hot and extremely cold
temperatures, and the damage (intact, minor damage, major damage) is recorded.
The researchers want to determine if the type of concrete has an association (a
relationship) with the type of damage. The results of the test are shown in the
following table. There are a total of 165 concrete samples.

Type A Type B Type C Type D

Intact 21 33 10 8
Minor damage 12 10 12 9
Major damage 2 1 22 25

The researchers carry out a chi-square test, and obtain the following output.
Balka ISE 1.09 13.4. CHAPTER EXERCISES 210

Pearson’s Chi-squared test

data: concrete
X-squared = 59.6202, df = 6, p-value = 5.376e-11

Of the following options, which one of the following is the best conclusion?

(a) There is very strong evidence that the 4 concrete types have the same distri-
bution of damage.
(b) There is little or no evidence of a di˙erence in distribution of damage among
the 4 concrete types.
(c) There is very strong evidence of a di˙erence in the distribution of damage
among the 4 concrete types.
(d) We can be very confdent that concrete type and damage are independent.
(e) There is not strong evidence of a di˙erence in population mean damage be-
tween the 4 concrete types.

29. A company is investigating new methods of eradicating bedbugs. They are testing
3 new methods: one based on heat, one based on chemicals, and a combination
heat/chemical method. Under controlled conditions, they subject 200 adult bed-
bugs to each of the 3 methods, and count the number of bedbugs that die. The
results are given in Table 13.16.

Heat Chemical Heat and Chemical

Dead 112 54 152
Alive 88 146 48

Table 13.16: Bedbugs killed by di˙erent methods.

Suppose we wish to test the null hypothesis that the 3 methods are all equally
e˙ective (the true proportion of bedbugs that would be killed under these conditions
is the same for all 3 methods). Carry out this test, giving the value of the test
statistic, p-value, and conclusion. (Hint to ease the calculation burden: the value
of the χ2 statistic is 97.47.)
Chapter 14

One-Way ANOVA

J.B.’s strongly suggested exercises: 2, 3, 5, 8, 10, 12, 13, 16, 18, 19, 22, 23, 25, 26, 28

14.1 Introduction

14.2 One-Way ANOVA

1. (a) In words and symbols, what is the null hypothesis in one-way ANOVA?
(b) In words and symbols, what is the alternative hypothesis in one-way ANOVA?

14.3 Carrying out the One-Way ANOVA

14.3.1 The Formulas

2. Researchers investigated the e˙ectiveness of four di˙erent exercise plans on weight

loss. Forty people of similar age and other characteristics were randomly assigned
to the di˙erent plans (10 to each plan). At the end of two months, weight loss was
measured. Consider the partially completed ANOVA table found in Table 14.1.

(a) Complete the ANOVA table (there is enough information to complete the
table).
(b) What is the value of the pooled variance?
(c) What is the value of the F statistic?

211
Balka ISE 1.09 14.3. CARRYING OUT THE ONE-WAY ANOVA 212

Source DF SS MS F p-value
Groups 800
Error — —
Total 1000 — — —

Table 14.1: A partially completed ANOVA table.

(d) Is there signifcant evidence at the 5% level that not all of these plans have
the same population mean weight loss?

3. The results from a three group experiment are found in Table 14.2.

Group 1 Group 2 Group 3

Sample mean 3.1 10.1 18.9
Sample standard deviation 2.1 4.2 3.7
Sample size 10 10 8

Table 14.2: Results from a three-group experiment.

Grand mean: X
¯ = 10.11429.

(a) What is the estimate of the within group variance (s2p )?

(b) What is the value of SST?
(c) Complete the entire ANOVA table.
(d) What is the p-value?
(e) Give an appropriate conclusion at the 5% level of signifcance.

4. Consider the following sample data.

Group A: 4, 6
Group B: 8, 12, 16
Group C: 4, 8

Suppose we wish to analyze this data with one-way ANOVA. (This might not be
the best idea, given the very small sample sizes, but suppose we are committed to
it.)

(a) What is the value of SST?

(b) What is the value of the pooled variance?
(c) What is the value of the F statistic?
Balka ISE 1.09 14.3. CARRYING OUT THE ONE-WAY ANOVA 213

14.3.2 An Example with Full Calculations

5. Walczyk et al. (2013) investigated possible di˙erences between truth tellers and
liars when questioned about a mock crime. Participants in a psychology experiment
were randomly assigned to a truth telling group, an unrehearsed lying group, or a
rehearsed lying group (where the individuals were allowed to see the questions and
think about their responses in advance). The participants were then asked several
questions. Some questions were general questions, in which all participants were to
answer truthfully, and some questions involved a mock crime, in which participants
were to either answer truthfully or lie, according to their assigned group. Various
characteristics of their responses were recorded.

In one aspect of the study, the researchers suspected that liars would tend to have
greater eye movement when compared to truth tellers. (They also suspected that
unrehearsed liars would have greater eye movement than rehearsed liars.) Table 14.3
illustrates the summary statistics of eye movement (in pixels) during the responses
to the general questions.

Group Truth tellers Unrehearsed lying Rehearsed lying

Mean 42 63 64
Standard deviation 36 44 38
Sample size 41 42 42

Table 14.3: Summary statistics of eye movement (in pixels).

(a) Is carrying out a one-way ANOVA appropriate here? How might one investi-
gate the ANOVA assumptions in greater detail?
(b) Plots of the eye movement measurements (not shown) show a little right skew-
ness. Is it still reasonable to use ANOVA here?
(c) The following output from R illustrates the results of the one-way ANOVA:
Df Sum Sq Mean Sq F value Pr(>F)
Groups 2 12757 6378 4.0866 0.01915
Residuals 122 190420 1561
Give a conclusion to the ANOVA F test at the 5% signifcance level.
(d) The ANOVA F test showed signifcant evidence against the null hypothesis
at the 5% signifcance level. Is the analysis complete? What else needs to be
done?
Balka ISE 1.09 14.4. WHAT SHOULD WE DO AFTER A ONE-WAY ANOVA? 214

14.4 What Should We Do After a One-Way ANOVA?

14.4.1 Introduction

6. Suppose we are about to draw 25 independent random samples and calculate 25

95% confdence intervals. Suppose that for this particular problem the intervals
can be considered independent, and the assumptions are perfectly justifed in each
case. What is the probability that all of the intervals capture the parameter that
they are estimating?

7. Suppose we are about to draw 10 independent random samples, and use the data
to test 10 null hypotheses, each at the 5% signifcance level. Suppose that for this
particular problem the tests can be considered independent, and the assumptions
of the test are perfectly justifed in each case. What is the probability that there
are no Type I errors made in the 10 tests?

14.4.2 Fisher’s LSD Method

8. Consider again the study frst discussed in Question 5, in which eye movement was
measured for truth tellers, unrehearsed liars, and rehearsed liars.

Group Truth tellers Unrehearsed lying Rehearsed lying

Mean 42 63 64
Standard deviation 36 44 38
Sample size 41 42 42

Table 14.4: Summary statistics of eye movement (in pixels).

Df Sum Sq Mean Sq F value Pr(>F)

Groups 2 12757 6378 4.0866 0.01915
Residuals 122 190420 1561

(a) How many pairwise comparisons are there?

(b) Use the LSD procedure to construct confdence intervals for the pairwise dif-
ferences. Use a confdence level of 95% for each of the individual intervals.
(c) Which intervals contain 0? Which of the observed di˙erences are statistically
signifcant at the 5% level?
(d) What is the family-wise confdence level?
(e) Give a summary of the results.
Balka ISE 1.09 14.4. WHAT SHOULD WE DO AFTER A ONE-WAY ANOVA? 215

14.4.3 The Bonferroni Correction

9. Suppose we run an experiment with 5 treatment groups and 6 observations in each

group. We carry out a one-way ANOVA, and fnd a p-value of 0.007. We then wish
to calculate confdence intervals for the pairwise comparisons using the Bonferroni
multiple comparison procedure.

(a) How many pairwise comparisons are there?

(b) If we want to keep the family-wise confdence level at at least 90%, what should
be the confdence level of the individual intervals?
(c) What is the appropriate t value for the calculation of the margin of error of
the intervals?

10. Consider again the study frst discussed in Question 5, in which eye movement was
measured for truth tellers, unrehearsed liars, and rehearsed liars.

Group Truth tellers Unrehearsed lying Rehearsed lying

Mean 42 63 64
Standard deviation 36 44 38
Sample size 41 42 42

Table 14.5: Summary statistics of eye movement (in pixels).

Df Sum Sq Mean Sq F value Pr(>F)

Groups 2 12757 6378 4.0866 0.01915
Residuals 122 190420 1561

(a) How many pairwise comparisons are there?

(b) Suppose we wish to use the Bonferroni procedure to construct confdence in-
tervals for the pairwise di˙erences. If we wish to have a family-wise confdence
level of 95%, what is the appropriate confdence level for the individual inter-
vals?
(c) Use the Bonferroni procedure to construct confdence intervals for the pairwise
di˙erences. (Helpful tip: Using software we can fnd that with 122 degrees of
freedom, t0.0083 = 2.4274.)
(d) Which intervals contain 0? Which of the observed di˙erences are statistically
signifcant at the 5% level?
(e) How do the Bonferroni intervals compare to the LSD intervals constructed in
Question 8?
(f) Give a summary of the results.
Balka ISE 1.09 14.5. EXAMPLES 216

14.4.4 The Tukey Procedure

11. Consider again the study frst discussed in Question 5, in which eye movement was
measured for truth tellers, unrehearsed liars, and rehearsed liars.

Group Truth tellers Unrehearsed lying Rehearsed lying

Mean 42 63 64
Standard deviation 36 44 38
Sample size 41 42 42

Table 14.6: Summary statistics of eye movement (in pixels).

Df Sum Sq Mean Sq F value Pr(>F)

Groups 2 12757 6378 4.0866 0.01915
Residuals 122 190420 1561

(a) Suppose we wish to use the Tukey procedure to calculate confdence inter-
vals for the pairwise di˙erences. The sample sizes are not equal. Is it still
reasonable to use the Tukey procedure?
(b) The results of the Tukey procedure at a family-wise confdence level of 95%
are summarized in the following output.
diff lwr upr p adj
TT-RL -22 -42.579482 -1.420518 0.0331275
UL-RL -1 -21.455134 19.455134 0.9926103
UL-TT 21 0.420518 41.579482 0.0443732
Verify that these are indeed the correct Tukey intervals by calculating them
yourself. (Helpful tip: Using software we can fnd that for 3 groups and 122
degrees of freedom the appropriate Tukey multiplier is 2.3725.)
(c) Which intervals contain 0? Which pairs of means have a statistically signifcant
di˙erence at a family-wise signifcance level of 5%?
(d) How do the Tukey intervals compare to the LSD intervals constructed in Ques-
tion 8 and the Bonferroni intervals constructed in Question 10?
(e) Give a summary of the results.

14.5 Examples

14.6 A Few More Points

12. What are the assumptions of one-way ANOVA? What are the consequences if these
assumptions are violated?
Balka ISE 1.09 14.7. CHAPTER EXERCISES 217

14.7 Chapter Exercises

14.7.1 Basic Calculations

13. (a) If we let F represent a random variable that has an F distribution with 5 and
8 degrees of freedom, what is P (F > 4.82)?
(b) If we let F represent a random variable that has an F distribution with 3 and
2 degrees of freedom, what is P (F < 19.16)?
(c) If we let F represent a random variable that has an F distribution with 6 and
4 degrees of freedom, what is the value a such that P (F > a) = 0.05?
(d) If we let F represent a random variable that has an F distribution with 12
and 27 degrees of freedom, what is the value a such that P (F > a) = 0.01?

14.7.2 Concepts

14. In a one-way ANOVA the F statistic is found to be 0.00002. Does this value give
strong evidence against the null hypothesis?

15. Consider a one-way ANOVA.

(a) Under what conditions would the test statistic be equal to 0?

(b) Under what conditions would the test statistic be equal to 1?

16. Consider the boxplots in Figure 14.1, representing 3 separate and independent sam-
ples of size 20.
20
18
16
14
12
10
8
A B C

Figure 14.1: Boxplots for 3 independent samples.

Which of the following statements are true?

(a) If we performed a one-way ANOVA on these 3 samples, then the p-value would
be small.
Balka ISE 1.09 14.7. CHAPTER EXERCISES 218

(b) If we carried out a t-test of H0 : µA = µC against a two-sided alternative, then

the p-value would be small.
(c) The use of the ANOVA procedures for the test of H0 : µA = µB = µC would
be a bad idea, as the assumptions are clearly violated.
(d) If we carried out a pooled-variance t-test of H0 : µA = µC against a two-sided
alternative, and a one-way ANOVA F test of H0 : µA = µC , then the test
statistics would have the relationship: t2 = F .
(e) If we carried out a pooled-variance t-test of H0 : µA = µC against a two-sided
alternative, then a one-way ANOVA of the test of H0 : µA = µC , the p-values
would be exactly equal.

17. Consider a one-way ANOVA, with 10 observations in each of 5 di˙erent groups.

Suppose the null hypothesis is true, and the assumptions are true.

(a) What is the distribution of the test statistic?

(b) What is the distribution of the p-value?
(c) What would the p-value equal on average?

18. Test your conceptual understanding: Which of the following statements are true?
You should be able to explain why a statement is true or why a statement is false.

(a) If the null hypothesis (and the assumptions) are true, then the test statistic
in one-way ANOVA has an F distribution.
(b) In one-way ANOVA, we assume that the observations within each group are
normally distributed, and that all groups have the same population variance.
(c) In one-way ANOVA, we assume that the observations within each group are
normally distributed, and that all groups have the same population mean.
(d) The test statistic in one-way ANOVA can be negative.
(e) If the null hypothesis is false, then MST will tend to be bigger than MSE.

19. Test your conceptual understanding: Which of the following statements are true?
You should be able to explain why a statement is true or why a statement is false.

(a) If the null hypothesis is true, then the F statistic will be infnite.
(b) If the null hypothesis is true, then the F statistic will equal 1.
(c) If the null hypothesis is false, then the F statistic will sometimes be less than
one.
(d) If the null hypothesis is false, then the F statistic will sometimes be less than
0.
(e) If the null hypothesis is false, then the p-value will be less than 0.05.

20. Test your conceptual understanding: Which of the following statements are true?
You should be able to explain why a statement is true or why a statement is false.
Balka ISE 1.09 14.7. CHAPTER EXERCISES 219

(a) If M SE = 0, then the p-value will equal the F statistic.

(b) If the population means are equal, then the F statistic will equal 1.
(c) If the sample means are all equal, then the F statistic will equal 1.
(d) If the sample means are all equal, then the null hypothesis is true.
(e) The mean square error is the pooled variance.

21. Test your conceptual understanding: Which of the following statements are true?
You should be able to explain why a statement is true or why a statement is false.

(a) If the population means are very di˙erent, we expect the F statistic to be
greater than one.
(b) We will reject the null hypothesis at the 5% level whenever the sample means
are all equal.
(c) An assumption of one-way ANOVA is that the groups all have di˙erent pop-
ulation variances.
(d) If the assumptions of the pooled-variance t procedures are met, then the t test
and the F test are equivalent tests, with t2 = F .
(e) The equal variance assumption is not important for large sample sizes, due to
the central limit theorem.

14.7.3 Applications

22. Researchers investigated the e˙ect of calorie-restricted diets on the longevity of

mice. They conducted an experiment, randomly assigning each of 331 mice to one
of 6 di˙erent diets:

• A control group that were allowed to eat an unlimited amount.

• R1, R2, R3, R4, and R5. These were 5 increasingly calorie restricted diets.

The mice were fed the diets until they died. The survival time, in months, was
recorded. The results are illustrated in Figure 14.2 and Table 14.7.1

Diet NP R1 R2 R3 R4 R5
Mean 27.4 32.7 42.3 42.9 39.7 45.1
Standard deviation 6.1 5.1 7.8 6.7 7.0 6.7
Sample size 49 57 71 56 49 49

Table 14.7: Survival time summary statistics for the six diets.
1
Data from Weindruch, R., Walford, R., Fligiel, S., and Guthrie, D. (1986). The retardation of aging
in mice by dietary restriction: longevity, cancer, immunity and lifetime energy intake. The Journal of
Nutrition, 116:641–654. The data used here is simulated data based on the summary statistics from this
study.
Balka ISE 1.09 14.7. CHAPTER EXERCISES 220

70
Survival Time (Months)
60
50
40
30
20

NP R1 R2 R3 R4 R5
Diet

Figure 14.2: Survival times for the di˙erent diets.

(a) Do the boxplots and summary statistics give any indication that the ANOVA
procedure would be inappropriate? (Do any of the assumptions appear to be
violated?)
(b) Consider the following output from the statistical software R:
Df Sum Sq Mean Sq F value Pr(>F)
Diets 5 12267.4 2453.5 55.079 < 2.2e-16
Residuals 325 14477.1 44.5
Give a proper interpretation of the results of the F test.
(c) There is extremely strong evidence that the groups do not all have the same
population mean lifetime. (In other words, there is very strong evidence that
the diets do not all have the same e˙ect.) Suppose we wish to investigate the
pairwise di˙erences using the LSD procedure. How many pairwise comparisons
are there?
(d) Use the LSD procedure to calculate a 95% confdence interval for the di˙erence
in means between the NP and R1 diets. Does the interval contain 0? Is
the observed di˙erence statistically signifcant at the 5% level? Give a non-
technical summary of the results of this comparison.
(e) Use the LSD procedure to calculate a 95% confdence interval for the di˙erence
in means between the R2 and R3 diets. Does the interval contain 0? Is
the observed di˙erence statistically signifcant at the 5% level? Give a non-
technical summary of the results of this comparison.

23. Researchers investigated a possible e˙ect of antivenom on swelling after injection of

a rattlesnake venom2 . In the experiment, 24 pigs were injected with Crotalus atrox
venom in the right hind leg, then they were randomly assigned to one of 3 groups:

• A group that received an intravenous antivenom infusion (IV group).

2
O˙erman et al. (2009). Subcutaneous crotaline Fab antivenom for the treatment of rattlesnake enven-
omation in a porcine model. Clinical Toxicology, 47:61–68.
Balka ISE 1.09 14.7. CHAPTER EXERCISES 221

• A group that received a subcutaneous injection of antivenom at the rattlesnake

envenomation site in addition to the IV (SC/IV group).

• A control group that received a saline solution (control group).

Figure 14.3 and Table 14.8 illustrate the change in leg volume after 8 hours.
350
Change in leg volume (mL)

300
250
200
150
100

Control IV SC/IV

Figure 14.3: Change in leg volume (mL) after 8 hours.

Treatment Control IV SC/IV

Mean 201.25 203.33 209.29
Standard deviation 112.63 56.18 45.04
Sample size 8 9 7

Table 14.8: Summary statistics of change in leg volume (mL) after 8 hours.

(a) Do the boxplots and summary statistics give any indication that the ANOVA
procedure would be inappropriate? (Do any of the assumptions appear to be
violated?)
(b) Consider the following output for the one-way ANOVA procedure:
Df Sum Sq Mean Sq F value Pr(>F)
Treatments 2 257 128 0.0214 0.9789
Residuals 21 126209 6010
The researchers suspected that the antivenom would reduce swelling. Does
this experiment give any evidence of this e˙ect?
(c) Are there any other statistical procedures that should be carried out after the
ANOVA?

24. A study investigated possible di˙erences in the e˙ect of di˙erent therapies on weight
gain in young women with anorexia. The young women received either no treatment
(Control), a Cognitive Behavioural treatment (CBT), or Family Therapy (FT). The
change in weight (kilograms) over the study period was recorded. Figure 14.4 and
Balka ISE 1.09 14.7. CHAPTER EXERCISES 222

Table 14.9 summarize the results.3

10
Weight change (kg)
5
0
-5

Control CBT FT

Figure 14.4: Change in body weight (kg) over the course of the study.

Treatment Control CBT FT

Mean −0.20 1.36 3.30
Standard deviation 3.62 3.32 3.25
Sample size 26 29 17

Table 14.9: Summary statistics of change in body weight (kg) over the course of the
study.

(a) Do the boxplots and summary statistics give any indication that the ANOVA
procedure would be inappropriate? (Do any of the assumptions appear to be
violated?)
(b) Consider the following output from the statistical software R:
Df Sum Sq Mean Sq F value Pr(>F)
Therapies 2 126.46 63.23 5.4223 0.006499
Residuals 69 804.62 11.66
Give an interpretation of the results of the F test.
(c) There is very strong evidence that the groups do not all have the same popu-
lation mean change in weight. (In other words, there is strong evidence that
the therapies do not all have the same e˙ect.) Suppose we wish to investi-
gate the pairwise di˙erences using the LSD procedure. How many pairwise
comparisons are there?
(d) Use the LSD procedure to calculate 95% confdence intervals for the pairwise
di˙erences in means. Which of the observed di˙erences are statistically signif-
icant at the 5% level? Give a non-technical summary of the results of these
comparisons.
(e) Use the Bonferroni procedure to calculate confdence intervals for the pair-
wise di˙erences in means, such that the family-wise confdence level is at least
95%. (Helpful tip: Using software we can fnd that with 69 degrees of freedom,
3
The data is taken from: Hand, D., Daly, F., Lunn, A., McConway, K., and Ostrowsky, E. (1994). A
Handbook of Small Data Sets. Chapman and Hall.
Balka ISE 1.09 14.7. CHAPTER EXERCISES 223

t0.0083 = 2.4537.) Which of the observed di˙erences are statistically signif-

cant? Give a non-technical summary of the results of these comparisons.

25. A study4 investigated Social Dominance Orientation (SDO) in a male inmate pop-
ulation in the UK. The article states that SDO is a “16-item scale [that] measures
a tendency to endorse the appropriateness and desirability of hierarchical relation-
ships.” (SDO tends to be high in individuals that believe it is right for certain
groups to dominate others.) Is there a di˙erence in the mean SDO of inmates of
di˙erent ethnic backgrounds?

Ethnicity White Asian/Asian British Black/Black British

Mean 50.89 49.13 53.87
Standard deviation 16.21 14.07 16.80
Sample size 328 23 18

Table 14.10: Summary statistics of SDO by ethnic background.

(a) Do the values given in Table 14.10 give any indication that ANOVA should
not be used? How could the assumptions of ANOVA be further investigated?
(b) The following output from R shows the results of the ANOVA calculations:
Df Sum Sq Mean Sq F value Pr(>F)
Ethnic groups 2 231 115 0.4439 0.6419
Residuals 366 95077 260
Give a conclusion to the ANOVA F test.

26. Consider again the study frst discussed in Question 25, which investigated Social
Dominance Orientation (SDO) in a male inmate population in the UK. In Ques-
tion 25 it was found that there was no evidence of a di˙erence in the mean SDO
scores for the di˙erent ethnic groups. The author of the paper decided to pool the
individuals together (ignoring the ethnic backgrounds) for the subsequent analysis.

(a) Suppose for a moment that there are meaningful di˙erences in the SDO be-
tween the ethnic backgrounds. If these di˙erences are ignored and the inmates
are pooled, how might that a˙ect the analysis?
(b) A major point of interest in the study was investigating possible di˙erences in
the SDO scores between di˙erent age groups. (Before the study, the author
hypothesized that younger inmates would tend to have higher SDO scores.) A
summary of SDO scores for di˙erent age groups is given in Table 14.11.
The following output from R shows the results of the ANOVA calculations:
Df Sum Sq Mean Sq F value Pr(>F)
Age groups 3 9265 3088 13.054 3.841e-08
Residuals 391 92497 237
4
Graham-Kevan, N. (2011). Investigating social dominance in a prison population. Journal of Criminal
Psychology, 1(1):15–23.
Balka ISE 1.09 14.7. CHAPTER EXERCISES 224

Age 18-25 26-35 36-45 45+

Mean 55.86 51.79 45.23 39.69
Standard deviation 14.90 15.89 15.65 14.67
Sample size 160 133 80 22

Table 14.11: Summary statistics of SDO by age group.

Give a conclusion to the ANOVA F test.

(c) Taking into account the results of the F test, and the summary statistics of
Table 14.11, give a non-technical summary of what this study tells us about
the relationship between SDO and age group.

14.7.4 Extra Practice Questions

27. Suppose an experiment investigated six di˙erent teaching methods. Twenty stu-
dents were assigned to each of the six teaching methods. After a certain period
of time in each program, the students took a test measuring their ability in the
subject. The researchers wanted to use one-way ANOVA to analyze the data. The
total sum of squares was found to be 8117.2. The best estimate of the within-group
variance was found to be 60.2. You have enough information to complete the full
ANOVA table.

(a) Complete the ANOVA table.

(b) What is the p-value of the test?
(c) Give an appropriate conclusion at the 5% level of signifcance.
(d) If we wished to use the LSD method to calculate 90% confdence intervals for
the pairwise di˙erences in the population means, what would be the appropri-
ate margin of error of the intervals?
(e) What assumptions are necessary for the methods used above to be reasonable?
How could we check to see if these assumptions are reasonable?

28. Researchers investigated the strength of 9 di˙erent types of seat belt strap. Random
samples of four straps of each type were obtained, and the breaking strength, in
thousands of Newtons, was measured for each strap. The research points of interest
was to test the null hypothesis that the belts all had the same average breaking
strength. If this hypothesis was rejected, the researchers wanted to investigate
which pairs of means were di˙erent. The researchers felt the assumptions of one-way
ANOVA are reasonable for this data. Consider the following partially-completed
ANOVA table.

Source DF SS MS F
Balka ISE 1.09 14.7. CHAPTER EXERCISES 225

Treatments
Error 12
Total 612

(a) Complete the ANOVA table.

(b) What is the p-value of the F test?
(c) Give an appropriate conclusion at the α = 0.05 signifcance level.
(d) Suppose we wish to calculate 95% confdence intervals for the di˙erences in
group means (µ1 − µ2 , µ1 − µ3 , etc.), using the Fisher LSD procedure. What
is the appropriate margin of error of the intervals?
(e) Which of the following statements are true?
i. The sample means were not all equal.
ii. This was a balanced experimental design (same sample size in each group).
iii. We know for certain the population mean breaking strengths of the dif-
ferent groups are not all equal.
iv. The pooled sample variance is equal to 12.

29. An experiment was performed to compare 5 di˙erent fuel additives for their e˙ec-
tiveness in improving gas mileage. A total of 20 new cars of the same make and
model were used, 4 for each fuel additive. The response variable was the number
of kilometres per litre of gas. A completely randomized design was used, and the
partially flled out ANOVA table is given in Table 14.12.

Source DF SS MS F
Treatments 680.2
Error
Total 958.2

Table 14.12: ANOVA table for the fuel additives question.

(a) In words and symbols, what are the hypotheses of the F test in the ANOVA
table?
(b) What assumptions are necessary in order for the one-way ANOVA methods to
be reasonable? How could we check to see if these assumptions are reasonable?
(c) Complete the ANOVA table.
(d) What is the p-value of the F test?
(e) Give an appropriate conclusion.

30. Researchers investigated the shear strength of 4 types of tin-lead solder. Each type
of solder was used on 5 solder joints (20 joints in total). To investigate possible
di˙erences in the mean shear strength between the di˙erent types of solder, the
researchers carried out a one-way ANOVA in the statistical package R, with the
following results.
Balka ISE 1.09 14.7. CHAPTER EXERCISES 226

Df Sum Sq Mean Sq F value Pr(>F)

types 3 168.016 56.005 12.657 0.0001709 ***
Residuals 16 70.796 4.425

(a) In words and symbols, what are the appropriate hypotheses of the F test in
the ANOVA table?
(b) Give an appropriate conclusion to the ANOVA F test.
(c) If we wanted to use Fisher’s LSD method to investigate possible di˙erences in
the means of the di˙erent groups, what is the appropriate 95% margin of error
of the individual intervals?
Chapter 15

Introduction to Simple Linear

Regression

J.B.’s strongly suggested exercises: 2, 4, 6, 8, 9, 12, 14, 15, 16, 18, 20, 37, 40, 41, 42, 43

15.1 Introduction

1. What is regression analysis?

2. Suppose we wish to use a student’s GPA in high school to help us predict their
GPA in frst-year university using a regression model.

(a) What would be the appropriate response variable? What would be the appro-
priate explanatory variable?
(b) If we wanted to improve the predictions of university GPA, what other ex-
planatory variables might we consider including in our model?

3. How does simple linear regression di˙er from multiple linear regression?

15.2 The Linear Regression Model

4. Consider the simple linear regression model: Y = β0 +β1 X +. Give interpretations
of all of the terms in the model.

5. What is the di˙erence in meaning between β̂1 and β1 ? What is the di˙erence in
meaning between β̂0 and β0 ?

227
Balka ISE 1.09 15.3. THE LEAST SQUARES REGRESSION MODEL 228

6. A sample of 30 (X, Y ) pairs is drawn, the values are plotted, and a well-ftting
line is drawn through the points. The results are illustrated in Figure 15.1. From
the plot, estimate the values of β̂0 and β̂1 . (In the future we will use software to
calculate these values for us, but we should be able to roughly estimate them from
a plot.)

^ =β
Y ^ +β
^ x
0 1
0
Y
-5

-10

2 4 6 8 10
X

Figure 15.1: A scatterplot with the ftted line Ŷ = β̂0 + β̂1 X. What are the values of β̂0
and β̂1 ?

15.3 The Least Squares Regression Model

7. How are the parameter estimates chosen in simple linear regression? What do we
call this method of parameter estimation?

8. Suppose in a simple linear regression scenario: ¯ )2 = 20.2, P(Yi − Y¯ )2 =

P
(Xi − X
P
6.5, (Xi − X)(Y¯ i − Y¯ ) = 10.1, Y¯ = 23.0, and X ¯ = 12.0.

(a) What are the values of β̂0 and β̂1 ?

(b) If the units of Y are metres, and the units of X are seconds, what are the
units of β̂0 and β̂1 ?

9. Qu et al. (2011) investigated various relationships between physical characteristics

of the lizard Phrynocephalus frontalis. In one part of the study, the researchers
investigated the relationship between the snout-vent length (mm) and abdomen
length (mm) of male lizards. (Part of the study involved investigating how these
relationships di˙ered between male and female lizards, but we will focus on just the
male lizards here.) The researchers regressed abdomen length (Y ) on snout-vent
length (X). The results are illustrated in Figure 15.2 and the following output.
Balka ISE 1.09 15.4. STATISTICAL INFERENCE IN SIMPLE LINEAR REGRESSION 229

Estimate Std. Error t value Pr(>|t|)

(Intercept) -5.73529 2.50608 -2.289 0.0331
snout_vent 0.63554 0.04847 13.113 2.79e-11

Abdomen length (mm) 30

28

26

24

46 48 50 52 54 56 58
Snout-vent length (mm)

Figure 15.2: Abdomen length vs snout-vent length for 22 male lizards (P. frontalis).

(a) Based on the output, what is the least squares regression line?
(b) Interpret β̂1 in the context of the problem at hand.
(c) Interpret β̂0 in the context of the problem at hand.
(d) Based on the regression line, predict the abdomen length for a male P. frontalis
lizard that has a snout-vent length of 52.0 mm.
(e) One of the lizards in the sample had a snout-vent length of 52.0 mm and an
abdomen length of 27.3 mm. What is the residual associated with this lizard?
(f) If we calculated the residuals for all 22 lizards in the sample and added them,
what value would we get?

15.4 Statistical Inference in Simple Linear Regression

15.4.1 Model Assumptions

10. What assumptions are required in simple linear regression in order to carry out
valid statistical inference procedures?

11. In a simple linear regression analysis with n = 7, the residuals were:

3.1, 2.4, −6.1, −.5, 8.9, −12.6, 4.8

(a) What is the best estimate of σ 2 (the variance about the regression line)?
(b) What is the sum of the residuals? Do they always sum to this value?
Balka ISE 1.09 15.5. CHECKING MODEL ASSUMPTIONS WITH RESIDUAL PLOTS 230

15.4.2 Statistical Inference for the Parameter β1

12. In practice, how often do we carry out a test of:

(a) H0 : β̂1 = 0?
(b) H0 : β0 = 0?
(c) H0 : β1 = 0?
(d) H0 : β1 = β0 ?

13. What is the sampling distribution of the sample slope? (Under the assumptions of
the linear regression model.)

14. Consider again the study frst discussed in Question 9, in which researchers investi-
gated the relationship between abdomen length (mm) and snout-vent length (mm)
in a type of lizard (P. frontalis). The following output summarizes the results of a
linear regression of abdomen length against snout-vent length for 22 male lizards.

Estimate Std. Error t value Pr(>|t|)

(Intercept) -5.73529 2.50608 -2.289 0.0331
snout_vent 0.63554 0.04847 13.113 2.79e-11

(a) Calculate a 95% confdence interval for β1 .

(b) Interpret the 95% confdence interval for β1 found in part 14a.
(c) Test the null hypothesis that there is no linear relationship between abdomen
length and snout-vent length for this type of lizard. Give appropriate hy-
potheses, test statistic, p-value and conclusion. (Although we would certainly
expect the abdomen length to increase with snout-vent length, carry out the
test using a two-sided alternative hypothesis.)
(d) Inference procedures on the intercept β0 proceed in a similar fashion to infer-
ence procedures on β1 , but they are not typically as interesting from a practical
viewpoint. Suppose in this question we wish to test H0 : β0 = 0 against a two-
sided alternative hypothesis. Carry out the test at a 5% signifcance level.
(e) Construct a 95% confdence interval for the parameter β0 .

15.5 Checking Model Assumptions with Residual Plots

15. Consider the residual plots given in Figure 15.3. (They represent residual plots for
4 di˙erent samples.) Would any of these plots indicate a problem with the assumed
linear regression model?

16. Consider the study that related abdominal length (mm) to snout-vent length (mm)
for P. frontalis lizards (frst discussed in Question 9). In Question 14 we carried
Balka ISE 1.09 15.5. CHECKING MODEL ASSUMPTIONS WITH RESIDUAL PLOTS 231

Residuals

Residuals
0
0

X X

(a) (b)

3
2
1
Sample Quantiles
Residuals

0
0

-1
-2
-3
-4
-2 -1 0 1 2
X Theoretical Quantiles

(c) (d) Normal QQ plot of the residuals.

Figure 15.3: The residual plots for Question 15.

out inference procedures on the slope and intercept. These inference procedures
required certain assumptions, and these assumptions should be checked by plotting
and looking at appropriate plots of the residuals. Figure 15.4 illustrates a plot of
the residuals versus snout-vent length, and a normal quantile-quantile plot of the
residuals. Do these residual plots indicate problems with the assumed linear model?
Were the inference procedures we conducted in Question 14 valid?

1
Residuals

Residuals

0
0

-1
-1

46 48 50 52 54 56 58 -1.5 -0.5 0.5 1.5

Snout-vent length Theoretical Quantiles

(a) Residuals vs snout-vent length. (b) Normal quantile-quantile plot of the

residuals.

Figure 15.4: Residuals vs snout-vent length and a normal quantile-quantile plot of the
residuals for Question 16.
Balka ISE 1.09 15.6. THE CORRELATION COEFFICIENT AND THE COEFFICIENT OF DETERMINATION 232

15.6 The Correlation Coeÿcient and the Coeÿcient of

Determination

17. Suppose in a simple linear regression scenario: ¯ 2 = 20.2, P(Yi − Y¯ )2 =

P
(Xi − X)
P ¯ i − Y¯ ) = 10.1, Y¯ = 23.0, and X
6.5, (Xi − X)(Y ¯ = 12.0.

(a) What is the value of r?

(b) If the units of Y are metres, and the units of X are seconds, what are the
units of the correlation coeÿcient?

18. Consider the plots in Figure 15.5.

(a) Would the correlation coeÿcient be a reasonable measure of the strength of

the relationship between Y and X in these scenarios? Justify your response.
(b) Rank the plots in order of increasing values of the correlation coeÿcient r.
(Also take an educated guess as to the value of r in each case.)
(c) Rank the plots in order of increasing strength of the linear relationship between
Y and X.
(d) The values of Y and X are not given on the axes in Figure 15.5. Would the
value of the correlation coeÿcient change if the scaling were changed on the
X and/or Y axes?

Y Y

X X

(a) (b)

Y Y

X X

(c) (d)

Figure 15.5: The plots for Question 18. Which plot has the largest value of the correlation
coeÿcient r?
Balka ISE 1.09 15.7. ESTIMATION AND PREDICTION USING THE FITTED LINE 233

19. Consider again the plots in Figure 15.5.

(a) Would the coeÿcient of determination R2 be a reasonable measure of the

strength of the relationship between Y and X in these scenarios? Justify your
response.
(b) Rank the plots in order of increasing values of the coeÿcient of determination
R2 . (Also take an educated guess as to the value of R2 in each case.)

20. Consider again the study frst discussed in Question 9, in which researchers investi-
gated the relationship between abdomen length (mm) and snout-vent length (mm)
in a type of lizard (P. frontalis). The following output summarizes the results of a
linear regression of abdomen length against snout-vent length for 22 male lizards.
(See Figure 15.2 for the scatterplot.)

Estimate Std. Error t value Pr(>|t|)

(Intercept) -5.73529 2.50608 -2.289 0.0331
snout_vent 0.63554 0.04847 13.113 2.79e-11
Residual standard error: 0.6218 on 20 degrees of freedom
Multiple R-squared: 0.8958, Adjusted R-squared: 0.8906

(a) What is the value of the coeÿcient of determination in given the output?
(Hint: It is called Multiple R-squared in the R output.)
(b) Give an interpretation of the value of R2 in the context of the problem at
hand.
(c) The correlation coeÿcient r is not given directly in the output, but it can be
found with a simple calculation. What is the value of the correlation coeÿcient
r?

15.7 Estimation and Prediction Using the Fitted Line

21. Consider again the study relating abdomen length (Y ) to snout-vent length (X) in
a type of lizard (P. frontalis). In Question 9 we found that the least squares ft
was: Ŷ = −5.73529 + 0.63554X. The mean snout-vent length for the 22 lizards in
the sample is 51.6 mm, and the mean abdomen length is 27.1 mm.

(a) What is the estimated mean abdomen length for male P. frontalis lizards that
have a snout-vent length of 50.0 mm?
(b) What is the predicted abdomen length for a male P. frontalis lizard that has
a snout-vent length of 50.0 mm?
(c) Suppose we wish to calculate a 95% confdence interval for the true mean ab-
domen length for lizards with a snout-vent length of 50.0 mm. Using software,
Balka ISE 1.09 15.8. TRANSFORMATIONS 234

we can fnd that the standard error of the estimated mean in this situation is
0.15447. Calculate the appropriate 95% confdence interval.
(d) Suppose we also wish to calculate a 95% confdence interval for the true mean
abdomen length for lizards with a snout-vent length of 45.0 mm. Would the
appropriate standard error be smaller or larger than the one in part 21c?
Would the confdence interval be wider or narrower than the one in part 21c?
(e) Suppose we wished to calculate a 95% prediction interval for the abdomen
length of a single lizard with a snout-vent length of 50 mm. Would this pre-
diction interval for a single value be wider, narrower, or the same width as the
confdence interval found in part 21c?

15.8 Transformations

15.9 A Complete Example

15.10 Outliers and Infuential Points

22. Consider the scatterplot given in Figure 15.6. Four of the points in this data set
are plotted in blue and labelled with letters. The regression line on the scatterplot
is from a regression with all points included.

B
D
Y
C
A

Figure 15.6: A sample data set.

(a) Which two points have the greatest leverage?

(b) Which two points have residuals with the greatest magnitude?
(c) Which point has the greatest infuence?
(d) If Point D were removed from the calculations, would the slope of the regression
line increase, decrease, or stay the same?
Balka ISE 1.09 15.11. SOME CAUTIONS ABOUT REGRESSION AND CORRELATION 235

15.11 Some Cautions about Regression and Correlation

15.11.1 Always Plot Your Data

15.11.2 Avoid Extrapolating

23. Consider again the study relating abdomen length (mm) to snout-vent length (mm)
in a type of lizard (P. frontalis). The results are illustrated in Figure 15.7 and the
following output.

Estimate Std. Error t value Pr(>|t|)

(Intercept) -5.73529 2.50608 -2.289 0.0331
snout_vent 0.63554 0.04847 13.113 2.79e-11

30
Abdomen length (mm)

28

26

24

46 48 50 52 54 56 58
Snout-vent length (mm)

Figure 15.7: Abdomen length vs snout-vent length for 22 male lizards (P. frontalis).

(a) Would it be reasonable to use the estimated regression line to predict the
abdomen length for a lizard with a snout-vent length of 35 mm?
(b) The value of β̂0 is given in the output as −5.73529. β̂0 is the estimated mean
of Y when X = 0. But the abdomen length cannot be negative, so this value
does not make practical sense. Is our model fawed?
Balka ISE 1.09 15.12. A BRIEF MULTIPLE REGRESSION EXAMPLE 236

15.11.3 Correlation Does Not Imply Causation

15.12 A Brief Multiple Regression Example

15.13 Chapter Exercises

15.13.1 Basic Calculations

24. Suppose X and Y measurements were taken on 10 individuals, with the results seen
in the following table.

Individual X Y
1 8 25
2 4 14
3 6 21
4 7 26
5 7 20
6 6 20
7 2 6
8 9 28
9 3 10
10 6 24
25
20
Y
15
10

2 3 4 5 6 7 8 9
X

Summary statistics: X
¯ = 5.8, Y¯ = 19.4, s2 = 5.15. The regression line is:

Ŷ = 1.20 + 3.14X and SE(β̂1 ) = 0.3436.

(a) What is the value of the residual for the frst individual?
(b) What is a 95% confdence interval for β1 ?
(c) Based on the 95% interval for β1 , is there strong evidence that X and Y are
linearly related?
(d) What is the predicted value of Y if X = 120? Is it reasonable to use the
regression line for this prediction?
(e) What is the predicted value of Y if X = 20? Is it reasonable to use the
regression line for this prediction?
Balka ISE 1.09 15.13. CHAPTER EXERCISES 237

25. Consider the following simple data set.

X 1 1 2 2 3
Y 3.1 2.7 1.1 1.7 .5

The following plot and output is a result of a simple linear regression of Y on X

for this data.

3.0

2.5

2.0
Y
1.5

1.0

0.5
Coefficients: 1.0 1.5 2.0 2.5 3.0
X

Value Std. Error t value Pr(>|t|)

(Intercept) 4.0571 0.4052 10.0138 0.0021
x -1.2429 0.2078 -5.9799 0.0094

Residual standard error: 0.3478 on 3 degrees of freedom

Multiple R-Squared: 0.9226
F-statistic: 35.76 on 1 and 3 degrees of freedom, the p-value is 0.009361

(a) If X = 1, what is the predicted value of Y based on the regression line?

(b) What is the residual associated with the frst value in the data set?
(c) Calculate a 90% confdence interval for the slope parameter.
(d) If there is truly no linear relationship between Y and X, what would the value
of the slope parameter be? What would the value of the intercept be?
(e) Suppose we wish to test the null hypothesis that the population slope is equal
to 2. What is the value of the appropriate test statistic? What is the p-value
of the test? What is an appropriate conclusion?
(f) What assumptions are necessary for the methods used above to be valid? Is
there any reason to believe the assumptions are violated in this case?

26. Consider the following R output from a simple linear regression.

Coefficients:
Value Std. Error t value Pr(>|t|)
(Intercept) -.6000 1.537 -.390 0.706
X 1.218 .248 4.919 0.001

(a) Is there strong evidence of a linear relationship between X and Y ?

(b) If the sample size is 10, what is a 95% confdence interval for β1 ?
Balka ISE 1.09 15.13. CHAPTER EXERCISES 238

15.13.2 Concepts

27. What is meant by the term extrapolation in simple linear regression?

28. Consider the three plots in Figure 15.8, representing simple linear regressions for 3
di˙erent data sets.

Figure 15.8: Three regression scenarios.

(a) Do any of the residual plots indicate a problem with the assumed model?
(b) For which data set is the strength of the relationship (not necessarily linear)
the strongest?
(c) For which data set is the strength of the linear relationship the greatest?
(d) The values of the coeÿcient of determination are 0.04392, 0.9509 and 0.5935.
(Not necessarily in order). Match the data set with its coeÿcient of determi-
nation.

29. If the correlation coeÿcient is very close to 0, does that imply that there is no
relationship between X and Y ?

30. If r2 = .95 is there a strong linear relationship between X and Y ? Do changes in

Balka ISE 1.09 15.13. CHAPTER EXERCISES 239

X cause changes in Y ?

31. The following output from R summarizes the results of a regression analysis with
n = 4. (It doesn’t usually make much sense to try to glean information from a
regression with such a small sample size, but this question illustrates the output
and some of the calculations.)

Value Std. Error t value Pr(>|t|)

(Intercept) -1.0000 2.1213 -0.4714 0.6838
x 5.1000 0.7746 XXXXX 0.0223

(a) What is the estimated regression line?

(b) Predict the value of Y if X = 2.5.
(c) Calculate a 95% confdence interval for the true slope.
(d) The test statistic for testing the hypothesis that β1 = 0 has been replaced by
“XXXXX” out. What is this missing value?
(e) If we did not have the p-value in the output, and had to look it up in the
t table, what would the appropriate range be for the p-value of the test of
H0 : β1 = 0 against a two-sided alternative?
(f) The values that were used to calculate the output were:
X 1 2 3 4
Y 4.1 8.2 16.3 18.4
Calculate the four residuals, and use these residuals to calculate s2 .

32. Suppose in a simple linear regression with n = 4, the frst 3 residuals are: 2.0,
3.4, -0.7, but you’ve forgotten to write the fourth residual down. What is the best
estimate of σ 2 , the variance about the regression line?

33. Suppose we randomly and independently sample 10 X values from the standard
normal distribution and 10 Y values from the standard normal distribution. We
pair them randomly into 10 (X, Y ) pairs, regress Y on X, and test H0 : β1 = 0 with
a t test.

(a) What is the distribution of the t test statistic?

(b) What is the average value of the t test statistic?
(c) What is the distribution of the p-value in this scenario?
(d) On average, what would be the value of the p-value?

34. What is the most common hypothesis test in simple linear regression scenarios?

35. Suppose a 95% confdence interval for β1 is found to be (1.1, 2.7). What does this
tell us about the p-value of the test of H0 : β1 = 0 against a two-sided alternative
hypothesis?
Balka ISE 1.09 15.13. CHAPTER EXERCISES 240

36. Test your conceptual understanding: Which of the following statements are true?
You should be able to explain why a statement is true or why a statement is false.

(a) The method used to fnd the sample slope and intercept is called the method
of least cubes.
(b) The residuals in a simple linear regression always sum to 0.
(c) If X¯ = 0, then β̂1 = 0.
(d) If Y = 0, then β̂0 = 0.
¯
(e) r has the same sign as β̂1

37. Test your conceptual understanding: Which of the following statements are true?
You should be able to explain why a statement is true or why a statement is false.

(a) β̂0 has the same sign as β̂1 .

(b) If all data points fall perfectly on a line, then r = 1 or −1.
(c) The least squares regression line always passes through the point (0, 0).
(d) The least squares regression line always passes through the point (X¯ , Y¯ ).
(e) If there is no linear relationship between Y and X, then r = 1.

38. Test your conceptual understanding: Which of the following statements are true?
You should be able to explain why a statement is true or why a statement is false.

(a) Tests and confdence intervals on β̂0 are the most common inference procedures
in regression.
(b) If the sample shows no linear relationship between Y and X (a random scat-
tering of points, say), then r and r2 will be close to 0.
(c) i = Yi − (β̂0 + β̂1 Xi ).
(d) The residuals (the ei values) are uncorrelated.
(e) The sample correlation coeÿcient is the proportion of the variance in Y that
is attributable to the linear relationship with X.

39. Test your conceptual understanding: Which of the following statements are true?
You should be able to explain why a statement is true or why a statement is false.

(a) Instead of least squares, it would be better to choose a regression line based
on minimizing the sum of the absolute value of the residuals, but statisticians
have not yet fgured out how to do so.
(b) The estimated standard deviation in a simple linear regression is the sum of
the squared residuals, divided by the sample size.
(c) If the test of the null hypothesis β1 = 0 results in a very small p-value, then
R2 must have been large.
(d) If s = 0 and β̂1 > 0, then r = 1.
Balka ISE 1.09 15.13. CHAPTER EXERCISES 241

15.13.3 Applications

40. Cook et al. (2012)1 investigated various aspects of larviposition of the blowfy Cal-
liphora varifrons. (Larviposition is the laying of larvae (rather than eggs) by insects
in which the eggs hatch inside the female.) In one aspect of the study, female fecun-
dity was investigated by trapping female C. varifrons blowfies in a feld, dissecting
them, and counting the number of live larvae they had inside. The researchers
wished to investigate whether there was a relationship between the size of the fe-
male (as measured by head width, in mm), and the number of live larvae they
carried. Figure 15.9 and the following output illustrate the results. (In Figure 15.9
the X values have been jittererd —a small random amount has been added to each
X. This makes it possible to distinguish two points that lie on top of one another.
Jittering is done for visual display only—the non-jittered values are used in the
regression calculations.)

Coefficients:
Estimate Std. Error t value Pr(>|t|)
(Intercept) -29.400 5.984 -4.913 1.13e-05
headwidth 17.004 1.636 10.396 9.01e-14

Residual standard error: 4.085 on 47 degrees of freedom

Multiple R-squared: 0.6969, Adjusted R-squared: 0.6905
F-statistic: 108.1 on 1 and 47 DF, p-value: 9.007e-14

50 10

45
Number of live larvae

5
40
Residuals

35
0
30

25
-5
20

15 -10
2.5 3.0 3.5 4.0 4.5 2.5 3.0 3.5 4.0 4.5
Female headwidth (mm) Female headwidth (mm)

(a) Scatterplot number of larvae against head (b) Plot of residuals against head width.
width.

Figure 15.9: Scatterplot of number of larvae against female head width and plot of the
residuals.

(a) Do these plots give any indication that the simple linear regression model is
not appropriate?
1
Cook et al. (2012). The laying of live larvae by the blowfy Calliphora varifrons (Diptera: Calliphoridae).
Forensic Science International, 223:44–66
Balka ISE 1.09 15.13. CHAPTER EXERCISES 242

(b) Can the assumptions of the simple linear regression model be true here? If
not, can we still use the model?
(c) What is the estimated relationship between the number of live larvae and
female head width?
(d) Give an interpretation of the sample slope β̂1 in the context of the problem at
hand.
(e) What is the estimate of the di˙erence in the mean number of live larvae carried
by female blowfies that di˙er in their head width by 0.20 mm?
(f) Is there strong evidence of a relationship between the number of live larvae
and female head width?
(g) What is the predicted number of live larvae for a female with a head width of
4.0 mm?
(h) What proportion of the variance in the number of live larvae can be explained
by the size of the female (as measured by head width)?

41. Fink et al. (2007) investigated a possible relationship between the hand grip strength
of young men and their facial attractiveness as perceived by young women. 32
male student volunteers had their hand grip strength measured in kilograms force
(kgf), and their facial attractiveness was assessed by 79 female students who viewed
pictures of the faces of the men on a computer. The facial attractiveness and hand
grip strength for the 32 men are illustrated in Figure 15.10. The following output
summarizes the results of the regression calculations.

Estimate Std. Error t value Pr(>|t|)

(Intercept) 1.88340 0.94977 1.983 0.0566 .
Grip_strength 0.01309 0.01845 0.709 0.4836

Residual standard error: 0.7952 on 30 degrees of freedom

Multiple R-squared: 0.0165, Adjusted R-squared: -0.01629
F-statistic: 0.5032 on 1 and 30 DF, p-value: 0.4836

(a) Is there strong evidence of a relationship between hand grip strength and
facial attractiveness? Can hand grip strength be used to predict a man’s facial
attractiveness? Carry out an appropriate hypothesis test, giving hypotheses,
the test statistic, p-value, and conclusion.
(b) For this sample data, what percentage of the variance in facial attractiveness
can be explained by the linear relationship with hand grip strength?

42. Cámara et al. (2012) investigated various aspects of energy expenditure and lactic
acid accumulation in elite cyclists. In one part of the study, the researchers in-
vestigated a possible relationship between a cyclist’s cadence (measured in rpm)
and their Gross Eÿciency at the onset of blood lactate accumulation (GE =
Power output
Energy expended × 100%). Each of the 12 elite cyclists in the study rode at their
Balka ISE 1.09 15.13. CHAPTER EXERCISES 243

Facial Attractiveness
4

30 40 50 60 70
Hand Grip Strength (kgf)

Figure 15.10: Facial attractiveness and hand grip strength for 32 young men.

preferred cadence, and the GE at the onset of blood lactate accumulation was mea-
sured. A scatterplot of GE versus cadence with the ftted regression line is given in
Figure 15.11a. Figure 15.11b shows the residuals of this ft. The R output for the
simple linear regression model is:

Coefficients:
Estimate Std. Error t value Pr(>|t|)
(Intercept) 46.24268 5.55656 8.322 8.32e-06
cadence -0.27713 0.06271 -4.419 0.00130
26

24
GE (%)

22

20

18

16
80 85 90 95 100
Cadence (rpm)

(a) Scatterplot of GE vs cadence.

1.5

1.0
Residuals

0.5

0.0

-0.5

-1.0

80 85 90 95 100
Cadence (rpm)

(b) Residuals vs cadence.

Figure 15.11: Scatterplot of GE vs cadence and plot of the residuals for the simple linear
regression model.
Balka ISE 1.09 15.13. CHAPTER EXERCISES 244

(a) Do the scatterplot and residual plot show any problems with the assumed
linear model? If so, how might we improve the model?
(b) The scatterplot and residual plot give some indication that a line may not ad-
equately model the relationship between cadence and GE. But assume for this
question that we feel the simple linear regression model provides a reasonable
ft. Do the results of the regression calculations given in the R output indicate
strong evidence of a relationship between GE and cadence?

43. Bremner et al. (2007) investigated a possible relationship between memory and
the volume of the right hippocampus (an area of the brain) in patients with post-
traumatic stress disorder (PTSD). Twenty-two veterans of the Vietnam war had the
volume of their right hippocampus measured by an MRI scanner. They were also
assigned a score on their verbal memory according to the Wechsler Memory Scale.
The following plot and output represent the results of a regression of Memory Scale
score on right hippocampus volume.

120
Memory Scale Score

100

80

60

40

20

0
800 900 1000 1100 1200 1300 1400 1500

Volume of Right Hippocampus (mm3)

Coefficients:
Estimate Std. Error t value Pr(>|t|)
(Intercept) -41.72277 29.37690 -1.420 0.17093
hipp_volume 0.08835 0.02440 3.621 0.00170

Residual standard error: 15.7 on 20 degrees of freedom

Multiple R-squared: 0.396, Adjusted R-squared: 0.3658
F-statistic: 13.11 on 1 and 20 DF, p-value: 0.001704

(a) What are the hypotheses of the hypothesis test that has a p-value of 0.17093?
Is this a meaningful test?
(b) What are the hypotheses of the hypothesis test that has a p-value of 0.00170?
Is this a meaningful test?
(c) Give a conclusion to the hypothesis test on the slope.
(d) What is a 95% confdence interval for β1 ?
(e) The frst observation in the data set had a right hippocampus volume of 955
mm3 and a Memory Scale score of 28. What is the value of the residual
Balka ISE 1.09 15.13. CHAPTER EXERCISES 245

associated with this value?

(f) What proportion of the variance in memory scores can be explained by the
linear relationship with right hippocampus volume?
(g) What is the value of the correlation coeÿcient?

44. A developer of solar panels is investigating the nature of the relationship between
the solar cell temperature rise above ambient (Y ) and the amount of insulation
(megawatts per square centimetre). They test 25 solar cells under a very specifc
set of conditions. The following scatterplot and output from the statistical package
R illustrate the results.
30
Temperature Rise Above Ambient

25
20
15
10

30 40 50 60
Insulation

Coefficients:
Estimate Std. Error t value Pr(>|t|)
(Intercept) 0.13215 1.51447 0.087 0.931
Insulation 0.36200 0.03211 11.275 7.59e-11

(a) What is the predicted value of temperature rise (Y ) for an insulation value of
44?

(b) Consider the test in the output has resulted in a p-value of 7.59e-11. Give the
hypotheses and an appropriate conclusion to this hypothesis test.

(c) Suppose that before this experiment began, based on theoretical arguments
the researchers felt that the slope of the true regression line would be 0.30.
Does this experiment provide any evidence against this hypothesis? Test the
null hypothesis that the population slope equals 0.30, against a two-sided
alternative hypothesis. Give appropriate hypotheses, value of the test statistic,
p-value, and conclusion at α = 0.05.

45. Suppose a Christmas tree grower wants to know the relationship between the age
of a tree and its height, for the species of trees that he grows. A random sample of
10 trees was drawn, and their age and height was recorded.

Some summary statistics:

¯ = 82, Y¯ = 260.
n = 10, SSXX = 917, SSY Y = 1519, SSXY = 1100, s2 = 24.935, X
Balka ISE 1.09 15.13. CHAPTER EXERCISES 246

Tree Age (months) Height (cm)

X Y
1 70 249
2 75 253
3 80 253
... ... ...
10 100 272

You should be able to determine that the least squares regression line is:

Ŷ = 161.6358 + 1.1996X

(a) When Y is regressed on X, what is the estimate of the slope?

(b) What is the standard error of the slope estimator?
(c) What is a 95% confdence interval for the population slope β1 ?
(d) Based on the above interval, is there strong evidence that X and Y are linearly
related?
(e) Test the null hypothesis that the population slope is 1, against the alternative
that it is di˙erent from 1. Give appropriate null and alternative hypotheses,
a test statistic, a p-value and a conclusion.
(f) What is the predicted height for a tree that is 92 months old?
(g) What is the estimate of the variance about the regression line?
(h) Given our usual assumptions, what is the sampling distribution of the sample
slope?
(i) What is the predicted height of a tree if X = −2? Does this prediction make
sense in this situation?
Bibliography

Anderson, T., Reid, D., and Beaton, G. (1972). Vitamin C and the common cold: a
double-blind trial. Canadian Medical Association Journal, 107:503–508.

Brabin et al. (2008). The second to fourth digit ratio (2d:4d) in women with and without
human papillomavirus and cervical dysplasia. American Journal of Human Biology,
20:337–341.

Bremner et al. (2007). MRI based measurement of hippocampal volume in patients

with combat-related posttraumatic stress disorder. American Journal of Psychiatry,
152(7):973–981.

Cámara et al. (2012). Infuence of pedaling technique on metabolic eÿciency in elite

cyclists. Biology of Sport, 29(3):229–233.

Chakhssi et al. (2010). Change during forensic treatment in psychopathic versus nonpsy-
chopathic o˙enders. 21:660–682.

Chase et al. (2007). Safety restraint injuries in fatal motor vehicle collisions. Forensic
Science, Medicine, and Pathology, 3:258–263.

Cook et al. (2012). The laying of live larvae by the blowfy Calliphora varifrons (Diptera:
Calliphoridae). Forensic Science International, 223:44–66.

Dale, R. (2010). Birth statistics for African (Loxodonta africana) and Asian (Elephas
maximus) elephants in human care: History and implications for elephant welfare. Zoo
Biology, 29:87–103.

Duran et al. (2013). Growth and weight status of rural texas school youth. American
Journal of Human Biology, 25:71–77.

Fayrouz et al. (2012). Relation between fngerprints and di˙erent blood groups. Journal
of Forensic and Legal Medicine, 19:18–21.

Fink et al. (2007). Male facial appearance signals physical strength to women. American
Journal of Human Biology, 19:82–87.

247
Balka ISE 1.09 BIBLIOGRAPHY 248

Franklin et al. (2012). Do ants need to be old and experienced to teach? The Journal of
Experimental Biology, 215:1287–1292.

Fryar et al. (2012). Anthropometric reference data for children and adults: United states,
2007–2010. National Center for Health Statistics. Vital Health Stat., 11(252).

Graham-Kevan, N. (2011). Investigating social dominance in a prison population. Journal

of Criminal Psychology, 1(1):15–23.

Guéguen et al. (2012). The sweet smell of. courtship: E˙ects of pleasant ambient fragrance
on womenÕs receptivity to a manÕs courtship request. Journal of Environmental
Psychology, 32:123–125.

Häkkänen et al. (2009). Gender di˙erences in Finnish homicide o˙ence characteristics.

Forensic Science International, 186:75–80.

Hand, D., Daly, F., Lunn, A., McConway, K., and Ostrowsky, E. (1994). A Handbook of
Small Data Sets. Chapman and Hall.

Harlioglu et al. (2012). An investigation on the sperm number and reproductive param-
eters of males in wild caught freshwater crayfsh (Astacus leptodactylus, eschscholtz).
Animal Biology, 62:409–418.

Heidal et al. (2012). Cost and calorie analysis of fast food consumption in college students.
Food and Nutrition Sciences, 3:942–946.

Holland et al. (2013). A magnetic pulse does not a˙ect homing pigeon navigation: a gps
tracking experiment. The Journal of Experimental Biology, 216:2192–2200.

Jauchem et al. (2013). E˙ects of a TASER conducted energy weapon on the circulating
red-blood-cell population and other factors in sus scrofa. Forensic Science, Medicine,
and Pathology, 9:308–320.

Koshy et al. (2010). Parental smoking and increased likelihood of female births. Annals
of Human Biology, 37(6):789–800.

Kuhn, E., Nie, C., O’Brien, M., Withers, R. L., Wintemute, G., and Hargarten, S. W.
(2002). Missing the target: a comparison of buyback and fatality related guns. Injury
Prevention, 8:143–146.

Leth (2009). Intimate partner homicide. Forensic Science, Medicine, and Pathology,
5:199–203.

Lorin de la Grandmaison et al. (2001). Organ weight in 684 adult autopsies: new tables
for a Caucasoid population. Forensic Science International, 119:149–154.

MacArthur, J. (1931). Linkage studies with the tomato. Transactions of the Royal
Canadian Institute, 18:1–19.
Balka ISE 1.09 BIBLIOGRAPHY 249

Mahdavi et al. (2013). Is there any relationship between leb antigen expression and
Helicobacter pylori infection? Blood Cells, Molecules and Diseases, 51:174–176.

Martin et al. (2015). Births: Final data for 2013. National vital statistics reports; vol 64
no 1.

Meinl et al. (2008). Comparison of the validity of three dental methods for the estimation
of age at death. Forensic Science International, 178:96–105.

Murata, Y. (2001). Studies on a novel bitter amino acid, pulcherrimine in the green sea
urchin gonads. PhD thesis, Kyoto University.

O˙erman et al. (2009). Subcutaneous crotaline Fab antivenom for the treatment of
rattlesnake envenomation in a porcine model. Clinical Toxicology, 47:61–68.

Petkovšek, S. and Pokorny, B. (2013). Lead and cadmium in mushrooms from the vicinity
of two large emission sources in Slovenia. Science of the Total Environment, 443:944–
954.

Porter et al. (2010). Evaluation of the e˙ectiveness of a brief deception detection training
program. The Journal of Forensic Psychiatry & Psychology, 21(1):66–76.

Qu et al. (2011). Sexual dimorphism and female reproduction in two sympatric toad-
headed lizards, Phrynocephalus frontalis and P. versicolor (Agamidae). Animal Biol-
ogy, 61:139–151.

Sforza et al. (2009). Age- and sex-related changes in the normal human ear. Forensic
Science International, 187:110.e1–110.e7.

Shalev (2011). Chronic food restriction augments the reinstatement of extinguished

heroin-seeking behavior in rats. Addiction Biology, 17:691–693.

Sosa et al. (2005). Life history of Megamelus scutellaris with description of immature
stages (Hemiptera: Delphacidae). Annals of the Entomological Society of America,
98(1):66–72.

Statistics Canada (2013). Canadian health measures survey: Cycle 2 data tables, 2009
to 2011. Statistics Canada Catalogue no. 82-626-X.

Stevenson et al. (2007). Attention defcit/hyperactivity disorder (ADHD) symptoms and

digit ratios in a college sample. American Journal of Human Biology, 19:41–50.

Tantius et al. (2014). Experimental studies on the tensile properties of human umbilical
cords. Forensic Science International, 236:16–21.

U.S. Department of Agriculture, Agricultural Research Service (2014). USDA national

nutrient database for standard reference, release 27. Nutrient Data Laboratory Home
Page, http://www.ars.usda.gov/ba/bhnrc/ndl.
Balka ISE 1.09 BIBLIOGRAPHY 250

Walczyk et al. (2013). Eye movements and other cognitive cues to rehearsed and un-
rehearsed deception when interrogated about a mock crime. Applied Psychology in
Criminal Justice, 29(1):1–22.

Weindruch, R., Walford, R., Fligiel, S., and Guthrie, D. (1986). The retardation of aging
in mice by dietary restriction: longevity, cancer, immunity and lifetime energy intake.
The Journal of Nutrition, 116:641–654.

Zorba et al. (2013). Evaluation of the accuracy of di˙erent molar teeth measurements in
assessing sex. Forensic Science, Medicine, and Pathology, 9:13–23.