Topic 1.
Estimation and confidence intervals
Point estimates and confidence intervals for the population mean
The analysis of point estimates and confidence intervals begins with the
computation of the population mean. We should consider two
possibilities:
1 The population standard deviation (σ ) is known.
2 The population standard deviation (σ ) is unknown. In this case, the
sample standard deviation (s) is an appropriate estimator for the
population standard deviation (σ ).
() Quantitative Methods February 2015 3 / 24
Topic 1. Estimation and confidence intervals
Point estimates and confidence intervals for the population mean.
Point estimates
Statistical computed through the sample information to estimate the
population parameter.
For instance, the sample mean, X , represent a punctual estimate to
the population mean, µ.
p, a sample proportion is a point estimator of π, that it is the
population proportion.
s, the sample standard deviation, is a point estimate of σ , that it is
the standard deviation.
Observation
A punctual estimator only provide a part of information. Although it is
expected that the point estimate approximates the population
parameter. It would be convenient to measure how distance there are
between two points.
A confidence interval
()
is used for this purpose. Quantitative Methods February 2015 4 / 24
Topic 1. Estimation and confidence intervals
Point estimates and confidence intervals for the population mean.
Confidence interval
Set of real numbers formed from a sample date so that there is the
possibility that the population parameter occurs within the called set
with a specified probability. We call to this probability confidence level.
() Quantitative Methods February 2015 5 / 24
Topic 1. Estimation and confidence intervals
Point estimates and confidence intervals for the population mean.
Example
For example, in a vineyard it is estimated that production of each strain
is reasonably close to a normal distribution with a standard deviation of
0.7 kg. A winemaker sampling performed on 10 strains and gets an
average of 5.5 kg per vine. We calculate a confidence interval for the
average production with a confidence level of 95 %
() Quantitative Methods February 2015 6 / 24
Topic 1. Estimation and confidence intervals
Point estimates and confidence intervals for the population mean.
Observation
When the sample is enough great, then the central limit theorem
provides the following affirmation:
99 % of sample means is situated to ±2,58 standard deviations of
the population mean.
The 95 % of samples means selected of the same population
belongs to ±1,96 standard deviations from population means, µ.
() Quantitative Methods February 2015 7 / 24
Topic 1. Estimation and confidence intervals
Point estimates and confidence intervals for the population mean
If we repeat infinite times the experiment in the same conditions, we
obtain that 95 % of the time, the population mean belongs into the
computed interval (−1,96, 1,96).
z A. mayor A. menor z A. mayor A. menor z A. mayor A. menor
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
1.41 0.9207 0.0793 1.95 0.9744 0.0256 2.49 0.9936 0.0064
1.42 0.9222 0.0778 1,96 0,975 0,025 2.50 0.9938 0.0062
1.43 0.9236 0.0764 1.97 0.9756 0.0244 2.51 0.994 0.006
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
() Quantitative Methods February 2015 8 / 24
Topic 1. Estimation and confidence intervals
Point estimates and confidence intervals for the population mean
Observation
Frequently, also we use another confidence levels, for instance 90 %. In
this case, the area between −z and z we need to be 0,90. To compute
the value of z it is necessary to see the normal table, and search the
head column that appear ’area mayor=0.90’, the file that correspond
with this number is 1,65 y 0,05.
() Quantitative Methods February 2015 10 / 24
Topic 1. Estimation and confidence intervals
Point estimates and confidence intervals for the population mean
t-distribution
The t-distribution is distribution of continuity probability, that it has
many similarities with the normal distribution.
X−µ
t= √
s/ n
The t-distribution is flatter and it is more large than the normal
distribution.
() Quantitative Methods February 2015 11 / 24
Topic 1. Estimation and confidence intervals
Point estimates and confidence intervals for the population mean
t-distribution t. Characteristics
The following characteristics are based that the t-distribution is
computed about a normal population or quasi normal population.
1 It is a continuous distribution.
2 The density function is like a bell and also is symmetric.
3
There is not a unique t-distribution, but a family of t-distributions. All
of them has the same mean 0, and their standard deviations differ
according the sample size, n.
4 If the sample size n is getting bigger, the t-distribution converges to
the normal distribution, due to the errors to compute s to estimate σ
decrease with bigger samples.
() Quantitative Methods February 2015 12 / 24
Topic 1. Estimation and confidence intervals
Point estimates and confidence intervals for the population mean
The t-distribution
For instance, we suppose that we take a sample in the Business
Faculty, and we observe that the number of absences for a month in a
sample of 10 students is 3, 8, 6, 8, 7, 6, 2, 7, 6, 1. If we suppose that
the number of absences has a normal distribution, . Compute the
confidence interval with 99 % confidence for the average of absences.
() Quantitative Methods February 2015 13 / 24
Topic 1. Estimation and confidence intervals
Point estimates and confidence intervals for the population mean
We suppose that the population distribution is normal.
X = 5, 4, s = 2, 5033 y n = 10.
Degrees of freedom are 10-1=9.
√
X ± ts/ n.
Confidence level 99 %,
In this case, we wish a confidence level of 99 %.
() Quantitative Methods February 2015 14 / 24
Topic 1. Estimation and confidence intervals
Point estimates and confidence intervals for the population mean
Intervalos de confianza
80 % 90 % 95 % 98 % ,99 %
Área de las dos colas
Grados de libertad .20 .10 .05 .0 2 ,01
1 3,078 6,314 12,706 31,821 63,657
2 1,886 2,920 4,303 6,965 9,925
3 1,638 2,353 3,182 4,541 5,841
4 1,533 2,132 2,776 3,747 4,604
5 1,476 2,015 2,571 3,365 4,032
6 1,440 1,943 2,447 3,143 3,707
7 1,415 1,895 2,365 2,998 3,499
8 1,397 1,856 2,306 2,896 3,355
9 1,383 1,833 2,262 2,821 3, 245
10 1,372 1,812 2,228 2,764 3,169
() Quantitative Methods February 2015 15 / 24
Topic 1. Estimation and confidence intervals
Point estimates and confidence intervals for the population mean
Example
s s 2, 5033 2, 5033
(X − t √ , X + t √ ) = (5, 4 − 3, 245 √ , 5, 4 + 3, 245 √ )=
n n 10 10
(2,8312, 7,9687)
() Quantitative Methods February 2015 16 / 24
Topic 1. Estimation and confidence intervals
Confidence interval of a proportion
Proportion
Quotient, percent that provides the sample part of a population that
has a particular characteristic.
We call p to the sample proportion, X the number of successes and
n the number of elements in the sample, then
X
p= .
n
Example
A recent survey indicated that 92 % of every 100 % respondents agreed
with daylight summer time to save energy. The proportion sample is
92/100, or 0,92, or 92 %.
() Quantitative Methods February 2015 17 / 24
Topic 1. Estimation and confidence intervals
Confidence level of a proportion
Population proportion
The population proportion is defined through π. This parameter
represents the number of population success. In this case, the
distribution function that represents this success is a binomimal
distribution.
1 The sample data are results of counts.
2 There are only two possibilities (success or failure).
3
For each experiment, the probability of a success is always the
same.
4 The experiment are independent. Namely, the result of a experiment
not influence in the next experiment.
The next values nπ and n(1 − π) must be greater or equal than 5. This
condition allows use the central limit theorem and we can use the
normal distribution to our study.
() Quantitative Methods February 2015 18 / 24
Topic 1. Estimation and confidence intervals
Confidence interval of a proportion
Confidence interval of a population proportion
r r !
p(1 − p) p(1 − p)
p−z ,p+z .
n n
() Quantitative Methods February 2015 19 / 24
Topic 1. Estimation and confidence intervals
Correction factor of a finite population
If we have study a population that it has a finite number of elements N
and we denote by n the number of elements into the sample, hen it is
necessary adjust the sample errors in the above formulas of confidence
intervals.
() Quantitative Methods February 2015 20 / 24
Topic 1. Estimation and confidence intervals
Correction factor of a finite population
The confidence interval to measure the mean through a finite
population if the standard deviation is unknown can be written by
r ! r !!
s N−n s N−n
X−t √ ,X+t √ .
n N−1 n N−1
() Quantitative Methods February 2015 21 / 24
Topic 1. Estimation and confidence intervals
Correction factor of a finite population
Example
There are 250 families in the Gomera island. A random sample of 40
families provide that the average annual contribution is 450 euros, and
the standard deviation is 75 euros. Can The population mean be 445 o
425 euros?
1 What is the average of the population? What is the best estimate of
the population average?
2 Analyze why you should use the correction factor for a finite
population.
3
Make a confidence interval of 90 % to the average population. What
are the extreme points to the confidence interval?
() Quantitative Methods February 2015 22 / 24
Topic 1. Estimation and confidence intervals
Correction factor of a finite population
Example
1 First of all we observe that the population is finite. 250 families.
2 We not known the population average, that it is the parameter that
we wish to known. The best estimate of the population average is
the population average of the sample, that it is 450 euros.
3
The sample is n/N = 40/250 = 0,16. We observe that the sample
represents 0.05 of the population, we need to use the formula FCP
to adjust the standard error in order to define the confidence interval.
4 X = 450, s = 75, N = 250, n = 40. We use the t-distribution with
39 = 40 − 1 grades of freedom for 90 % is 1,685.
5
r !
s N −n
X ±t √ ⇒ (431,65, 468,53)
n N −1
() Quantitative Methods February 2015 23 / 24
Topic 1. Estimation and confidence intervals
Choosing the suitable size of a sample
Frequently, a very important question when we design a statistic study
is to determine the sample size, namely, how many elements has a
sample. The suitable sample size depends of three points:
1 The confidence level desired.
2 The error that we are willing to tolerate, namely the different
between zero and one of extreme point of the confidence interval.
3
The variability of the population under study. If the population is very
dispersed then we need a great sample size. On the other hand, if
the population are concentrated the the size of the sample must be
smaller.
() Quantitative Methods February 2015 24 / 24
Topic 1. Estimation and confidence intervals
Choosing the suitable size of a sample
Sample size to estimate the population mean
zσ 2
n=
E
1 n is the sample size.
2 z is the normal value that corresponds to the confidence interval
desired.
3
σ is the standard population deviation.
4 E is the admissible maximum error.
() Quantitative Methods February 2015 25 / 24
Topic 1. Estimation and confidence intervals
Choosing the suitable size of a sample
In some cases we don’t have the p population proportion. Then we take
p = 0, 50. Due to the formula
n = p(1 − p)(z/E)2
we observe that the p(1 − p) polynomial has a maximum point in
p = 0, 5. Namely, for every p ∈ [0, 1] and we obtain that
p(1 − p) < 0, 50(1 − 0, 50) = 0, 25
() Quantitative Methods February 2015 27 / 24
