Business Decision Making II
Interval Estimation
Dr. Nguyen Ngoc Phan
Dr. Nguyen Ngoc Phan Interval Estimation
Confidence Intervals for Means
We often here claims about population means.
Example: It is claimed that the mean protein content of a liter
carton of milk is 39 grams. The truth of this claim could only
be known by measuring the protein content of every 1 liter
carton of milk. This is clearly an impossible task. However, it
is possible to draw reasonable conclusions from measuring the
protein content of a random selection of cartons.
Suppose we randomly select a sample of milk cartons, and the
mean protein content for the sample is x̄ = 38 g. This could
be seen as evidence to dispute the claim that µ = 39 g.
However, it is possible that the average protein content is
indeed 39 g, and that the sampled cartons had lower than
average protein.
Dr. Nguyen Ngoc Phan Interval Estimation
Our sample cannot tell us with certainty the exact population
mean µ. However, we can use it to predict an interval in which
we are confident the population mean will lie. This interval is
called a confidence interval.
Definition
A confidence interval for a population mean µ, is an interval of
values between two limits, together with a percentage
indicating our confidence that µ lies in that interval.
Consider the statement ”We are 95% confident that the mean
protein content of milk cartons is between 37 g and 40 g”.
The statement indicates that the probability that the interval
[37, 40] grams contains the population mean is 0.95.
Dr. Nguyen Ngoc Phan Interval Estimation
Consider the distribution Xn of all sample means x̄ from large
sample of size n taken from population X with population
mean µ and standard deviation σ. Xn is√normally distributed
with mean µ and standard deviation σ/ n.
The corresponding standard normal random variable is
Xn − µ
Z= √
σ/ n
and Z ∼ N(0, 12 ).
For a 95% confidence interval we need to find a value a for
which P(−a ≤ Z ≤ a) = 0.95.
Using technology, we find that a ≈ 1.96.
Dr. Nguyen Ngoc Phan Interval Estimation
95% Confidence interval
The 95% confidence interval for µ given data set with sample
mean x̄ is
σ σ
x̄ − 1.96 √ ≤ µ ≤ x̄ + 1.96 √ ,
n n
which can also be written as
σ σ σ
x̄ ± 1.96 √ or x̄ − 1.96 √ , x̄ + 1.96 √ .
n n n
Dr. Nguyen Ngoc Phan Interval Estimation
Other Confidence Interval for µ
If we want a confidence interval that contains µ with
probability 1 − α, we find a value a such that
α
P(−a ≤ Z ≤ a) = 1 − α or equivalently P(Z ≤ −a) = .
2
The particular value of a we seek is therefore labeled by z α2 .
Dr. Nguyen Ngoc Phan Interval Estimation
General confidence interval for µ
The general confidence interval for µ given a data set with
sample mean x̄ is
σ σ
x̄ − z α2 √ ≤ µ ≤ x̄ + z α2 √
n n
The interval contains µ with probability 1 − α, so the
confidence level of the interval is (1 − α) × 100%.
Notice that:
We can determine the value of z α2 for any α using
technology.
The margin of error is z α2 √σn . The width of the confidence
interval is 2 × z α2 √σn .
Dr. Nguyen Ngoc Phan Interval Estimation
Some popular confidence level:
Confidence level α z α2
90% 0.1 1.645
95% 0.05 1.960
98% 0.02 2.326
99% 0.01 2.576
Example: Problem 10, page 353.
Dr. Nguyen Ngoc Phan Interval Estimation
Interval estimation using Excel: σ known
The sample mean is calculated by =AVERAGE(Input
range).
The margin of error is calculated by
=CONFIDENCE.NORM(α,σ,n).
Dr. Nguyen Ngoc Phan Interval Estimation
σ Unknown
When developing an interval estimate for a population mean
we usually do not have a good estimate of the population
standard deviation either. In these cases, we must use the
same sample to estimate both µ and σ. This situation
represents σ unknown.
When the sample standard population s is used to estimate σ,
the interval estimate for the population mean is based on the t
distribution.
Dr. Nguyen Ngoc Phan Interval Estimation
Interval estimate of a population mean: σ unknown
s
x̄ ± tα/2 √
n
where s is the sample standard deviation, (1 − α) is the
confidence coefficient, and tα/2 is the t value determined by t
distribution with n − 1 degree of freedom.
In most applications, a sample size of n ≥ 30 is adequate to
use this expression.
Example: The credit card balances, in page 357. Problem 20
in page 362.
Homework: Problems 17 and 21, page 363.
Dr. Nguyen Ngoc Phan Interval Estimation
Determining the sample size
When using confidence intervals to estimate population
means, we generally want a narrow confidence interval.
For a given confidence level, we can decrease the width of the
confidence interval by increasing the sample size.
When we wish to estimate the population mean to lie with
95% confident within an interval of width w , the sample size
required is 2
2 × 1.96σ
n=
w
where σ is the population standard deviation.
For other confidence levels, we can simply replace 1.96 by the
corresponding value of zα/2 .
Dr. Nguyen Ngoc Phan Interval Estimation
Example: Problem 26, page 365.
Homework: Problems 29 and 30, page 366.
Dr. Nguyen Ngoc Phan Interval Estimation
Confidence Intervals for Proportions
In general, the true population proportion p is unknown and
the purpose of taking a sample is to obtain a sample
proportion p̄ to estimate p.
For a sample of size n with sample proportion p̄, the 95%
confidence interval for p is
r
p̄(1 − p̄)
p̄ ± 1.96 .
n
For other confidence level we simply replace 1.96 by the
corresponding value of zα/2 .
Dr. Nguyen Ngoc Phan Interval Estimation
When we estimate a proportion it is desirable to make the
width of the confidence interval as small as possible. We can
make the width smaller by selecting a larger sample size n.
Choosing the sample size
For a 95% confidence interval of width w and preliminary
proportion p̄ = p ∗ , the sample size needed is
2
2 × 1.96
n= p ∗ (1 − p ∗ ).
w
If we have no preliminary value for p̄, then we assume the
worst case scenario, and choose the value p ∗ which maximizes
the value of p ∗ (1 − p ∗ ). This value is p ∗ = 0.5.
Dr. Nguyen Ngoc Phan Interval Estimation
Example: Problems 38 and 43, page 371
Homework: Problems 41 and 42, page 371.
Dr. Nguyen Ngoc Phan Interval Estimation
