Lecture notes for Design and analysis of experiments (Stat 2043) Chapter - 2

Chapter - Two

Simple Comparative Experiments

Contents

2. Simple comparative experiments.

2.1. Inference about the Difference in Means.

2.1.1. Independent samples.

2.1.2. Paired Sample t-Test.

2.2. Inference about the Variance of Normal Distribution.

Simple Comparative experiments

“Absence of evidence is not evidence of absence”

In this chapter, we consider experiments to compare two conditions (sometimes called

treatments). These are often called simple comparative experiments.
 Inference is the process of making interpretations or conclusions from sample data for the totality of
the population.
 It is only the sample data that is ready for inference.
 In statistics there are two ways though which inference can be made.
Statistical estimation
Statistical hypothesis testing.
Statistical Estimation

This is one way of making inference about the population parameter where the investigator does
not have any prior notion about values or characteristics of the population parameter.
Estimation: is a process by which we estimate the unknown population parameter from sample statistic.

Estimator: is any statistic that is used to estimate a population parameter.

[email protected]
10
Lecture notes for Design and analysis of experiments (Stat 2043) Chapter - 2

Estimate: is a numerical value of an estimator is called an estimate.

There are two ways of statistical estimation.

Point Estimation

o It is a procedure that results in a single value as an estimate for a parameter.

o A single value used to estimate a parameter.
 Interval Estimation
It is the procedure that results in the interval of values as an estimate for a parameter, which is
interval that contains the likely values of a parameter.
Interval Estimation: A range of values used to estimate a parameter.
Definitions

Confidence Interval: An interval estimate with a specific level of confidence.

Confidence Level: The percent of the time the true value will lie in the interval estimate given.

Degrees of Freedom: The number of data values which are allowed to vary once a statistic has been
determined.

o Properties of best estimator

Unbiased Estimator: An estimator whose expected value is the value of the parameter
being estimated. i.e., E( ̂) = θ.
Consistent Estimator: An estimator which gets closer to the value of the parameter as the
sample size increases. i.e., ̂ gets closer to θ as the sample size increases.
Relatively Efficient Estimator: The estimator for a parameter with the smallest variance.

Hypothesis Testing

- This is also one way of making inference about population parameter, where the investigator has prior
notion about the value of the parameter.
Definitions:

- Statistical hypothesis: is an assertion or statement about the population whose plausibility is to be

evaluated on the basis of the sample data.

[email protected]
11
Lecture notes for Design and analysis of experiments (Stat 2043) Chapter - 2

- Test statistic: is a statistics whose value serves to determine whether to reject or accept the
hypothesis to be tested.
- Statistic test: is a test or procedure used to evaluate a statistical hypothesis and its value depends on
sample data.
 There are two types of Hypothesis:

Null hypothesis:

- It is the hypothesis to be tested.

- It is the hypothesis of equality or the hypothesis of no difference.
- Usually denoted by H0.
Alternative hypothesis:

- It is the hypothesis available when the null hypothesis has to be rejected.

- It is the hypothesis of difference.
- Usually denoted by H1 or Ha.
Types of errors:

- The following table gives a summary of possible results of any hypothesis test:

Decision

Reject H0 Don't reject H0

H0 is true Type I Error Right Decision

H0 is false Right Decision Type II Error

Type I error: Rejecting the null hypothesis when it is true.

Type II error: Failing to reject (Don, t reject) the null hypothesis when it is false.
The value 1 − is called the power of the test. It represents the probability of not making
a Type II error.

[email protected]
12
Lecture notes for Design and analysis of experiments (Stat 2043) Chapter - 2

General steps in any Hypothesis testing:

1. First specify the null hypothesis (H0) and the alternative hypothesis (H1).

2. Identify / select a significance level,  .

3. Identify the sampling distribution of the estimator (Z – test, t - test, chi –

square, F – distribution… etc. ).

4. Calculate / Determine a test statistic .

5. Identify the critical region/Value.

6. Making decision.

7. Making conclusion.

2.1 Inference about the Difference in Means

When we deal with the inference about the difference in means, two things will come into our
mind. The first is independent sample and the second is paired sample. In the case of paired
sample, the sample is selected from a single population and these sample elements may be
treated at two different circumstances at different period of time such as pre and post, before and
after and the like. Moreover, the sample size at two different situations and the subjects are the
same.
In the case of independent samples, the samples are selected from different or the same
population independently. That is, the subjects are different and the sample sizes may or may not
be the same.
2.1.1 independent samples
Comparative studies are designed to discover and evaluate difference between treatments
(groups). In situations where we are making inferences about µ1 - µ2 based on random samples
independently selected from two populations to make inference on the difference between two
populations mean. The population and sample size may not be equal.

Case One: Consider the situation in which we are independently selecting random samples from
two populations that have normal distributions with different means µ1 and µ2 and population
standard deviations δ1 and δ2, are known (If the two populations variance are known), whatever

[email protected]
13
Lecture notes for Design and analysis of experiments (Stat 2043) Chapter - 2

the sample size (i.e. it may be less than 30, or greater than or equal to 30), then the appropriate
test statistics is.
(̅ ̅ ) ( )
Zcal = , If we have the same standard deviation on the two population, then the
√

(̅ ̅ ) ( )
test statistics becomes: Zcal =
√

The hypothesis test statistics about the difference between two population means. As with any
test procedure, we begin by specifying a hypothesis for the difference in population means.
Then one can formulate the hypothesis as follows
Hypothesis Test Statistics Criteria for Rejection
Reject H0 if | |> Zα/2
(̅ ̅ ) ( )
Zcal = Reject H0 if Zcal > Zα
√

Reject H0 if Zcal < -Zα

A 100 (1 - α) % confidence interval for the difference in mean for the two independent
samples when the population variances are known whatever the sample size is:

(̅ ̅ ) Zα/2 √ (̅ ̅ ) Zα/2 √ , Or

(̅ ̅ ) ± Zα/2 √

Example: To compare average score of urban and rural students’ independent samples

of n1  10 for urban, n2  10 for rural applicants to a college were selected. The sample

mean scores were calculated to be 16.015 and 16.005 for urban and rural respectively.

Past appearance shows that the admission test scores for both urban and rural students

have the standard deviation of and respectively.

A, At α = 0.05, can it be concluded that there is a significant difference between the

mean scores of urban and rural students?

B, Find a 95 percent confidence interval on the difference in the mean of the urban and

rural students.

[email protected]
14
Lecture notes for Design and analysis of experiments (Stat 2043) Chapter - 2

Solution:-

Given

Urban Rural

A,

Step 1: versus
Step 2:   0.05
Step 3: Identify the test statistic
Z – Test Because, The Population standard deviation is Known.
Step 4. Determine the test Statistic

(̅ ̅ ) ( ) ( ) ( )
Zcal =
√ √

Step 5: Identify the critical region, Zα/2 = Z0.05/2 = Z0.025 =

Step 6: Make Decision, Reject H0 if | | > Zα/2

1.35 < 1.96, So Don’t Reject H0 (fail to reject H0 )

Step 7: Conclusion, at 5% level of significance we have enough evidence / sufficient evidence to

conclude that the mean scores of the urban and rural students are equal.

[email protected]
15
Lecture notes for Design and analysis of experiments (Stat 2043) Chapter - 2

B,

The 95% confidence interval is

(̅ ̅ ) Zα/2 √ (̅ ̅ ) Zα/2 √

( ) 1.96 √ ( ) 1.96 √

-0.0045 = (- 0.0045, 0.0245)

Conclusion: This implies at 5% level of significance, we have enough evidence to conclude that
there is no difference (equal) between scores mean of urban and rural students.
Or we have 95% confident that the difference scores mean of urban and rural students are found
or lie between -0.0045 and 0.0245.

Case Two: Consider the situation in which we are independently selecting random samples from
two populations that have normal distributions with different means µ1 and µ2 and standard
deviations δ1 and δ2 are unknown (Population variance are unknown) that means sample variance
is known but large sample size (n ), the appropriate test statistics is Z - test statistics.

The test statistics is:

(̅ ̅ ) ( )
Zcal =
√

Then one can formulate the hypothesis as follows

Hypothesis Test Statistics Criteria for Rejection

Reject H0 if | | > Zα/2
(̅ ̅ ) ( )
Zcal = Reject H0 if Zcal > Zα
√

Reject H0 if Zcal < -Zα

[email protected]
16
Lecture notes for Design and analysis of experiments (Stat 2043) Chapter - 2

 A 100 (1-α) % confidence interval for the difference in mean for the two independent
samples when the sample variances are known with large sample size is:

(̅ ̅ ) Zα/2 √ (̅ ̅ ) Zα/2 √

Or

̅ ̅ ± Zα/2 √

Example. A consumer group selected independent random samples of supper - markets

located throughout a country for the purpose of comparing the retail prices per pound

of coffee of brands A and B. The results of the investigation are summarized below.

n1 = 75 n2 = 64
̅ =3 ̅ = 2.95
S1 = 0.11 S2 = 0.09
A, Test the hypothesis that the mean retail price per pound of brand A coffee is

significantly higher than the mean retail price per pound of brand B coffee? Use a level

of significance α = 0.01.

B, Find a 99 percent confidence interval on the difference in the mean retail price per

pound of brand A coffee and in the mean retail price per pound of brand B coffee.

Solution:-

Given
n1 = 75 n2 = 64
̅ =3 ̅ = 2.95
S1 = 0.11 S2 = 0.09

[email protected]
17
Lecture notes for Design and analysis of experiments (Stat 2043) Chapter - 2

A,

Step 1: H0: µ1 - µ2 = 0 (µ1 = µ2) (i.e., no difference between mean retail prices for brand A and B)

H1: µ1 - µ2 > 0 (µ1 > µ2) (i.e., mean retail price per pound of brand A is higher than that of
brand B).

Where: µ1= Mean retail price per pound of brand A coffee at all super-markets.

µ2= Mean retail price per pound of brand B coffee at all super-markets.

Step2: Identify the significance level, α = 0.01

Step 3: Identify the test statistic

Z – Test Because, the sample is large .
Step 4. Determine the test Statistic
(̅ ̅ ) ( ) ( ) ( )
Zcal =
√ √

Step 5: Identify the critical region, Zα = Z0.01= 2.33

Step 6: Make Decision, Reject H0 if Zcal > Zα

2.947 > 2.33, So Reject H0 ( Don’t reject H1)

Step 7: Conclusion, At 1% level of significance we have enough evidence / Sufficient evidence

to conclude that the mean retail price per pound of brand A coffee is significantly higher than the
mean retail price per pound of brand B coffee.

B, The 99% confidence interval is

NB: Zα/2 = =

(̅ ̅ ) Zα/2 √ (̅ ̅ ) Zα/2 √

(3 - 2.95) – 2.58√ (3 - 2.95) + 2.58 √

0.0062 = (0.0062, 0.0938)

[email protected]
18
Lecture notes for Design and analysis of experiments (Stat 2043) Chapter - 2

Conclusion: This implies at 1% level of significance we have enough evidence / Sufficient

evidence to conclude that that the mean retail price per pound of brand A coffee is significantly
higher than the mean retail price per pound of brand B coffee.

Or we have 99% confident that the mean difference for retail price per pound of brand A coffee
and the mean retail price per pound of brand B coffee is lie between 0.0062 and 0.0938.

Case Three: If the two population variance is assumed to be equal ( 12   22 ), but unknown and

n1 and n2 are small ( ), then we estimate  12 and  22 by S12 and S 22 .

In this case: - The test statistics becomes:

(̅ ̅ ) ( )

If two population variances are assumed to be equal but, the common variance  2 is unknown
we have to estimate  2 by pooled sample variance S p2 .

( ) ( )
= Pooled Variance,

Where are the variance of the first group and the Variance of the second group
respectively.

=√ = Pooled Standard Deviation

Then one can formulate the hypothesis as follows

Hypothesis Test Statistics Criteria for Rejection

( )
Reject H0 if | | ⁄

(̅ ̅ ) ( ) ( )
Reject H0 if
√ ( )
Reject H0 if

[email protected]
19
Lecture notes for Design and analysis of experiments (Stat 2043) Chapter - 2

 A 100(1 − )% confidence interval for the difference in mean for the two independent
samples if is :-

( ) ( )
(̅ ̅ ) ⁄ √ (̅ ̅ ) ⁄ √

( ) ( )
(( ̅ ̅ ) ⁄ √ (̅ ̅ ) ⁄ √ )

Example, A new filtering device is installed in a chemical unit; Before its installation
a random sample yielded the following information about the percentage of impurity:

and after installation a random sample

yielded the following information

Assume the population variance are assumed to be equal ( ) but unknown then

A, Test the hypothesis at 0.05, is there a significant difference in the filtering device

percentage of impurity.

B, Make the 95% confidence interval in the mean difference of filtering device percentage of

impurity.

Solution:-

Given

[email protected]
20
Lecture notes for Design and analysis of experiments (Stat 2043) Chapter - 2

A, Step1: versus
Step 2:   0.05
Step 3: Identify the test statistic
t – Test, because, the population variance is assumed to be equal and sample size is small.
Step 4. Determine the test Statistic

(̅ ̅ ) ( ) ( ) ( )

√ √

( )
Step 5: Identify the critical region, ⁄ = 2.131

( )
Step 6: Make Decision, Reject H0 if | | ⁄

| | < 2.131, So Don’t Reject H0

Step 7: Conclusion, at 5% level of significance we have enough evidence / sufficient evidence to

conclude that the mean of new filtering device before installation and after installation is the
same.

B, The 95% confidence interval is :

( ) ( )
(̅ ̅ ) ⁄ √ (̅ ̅ ) ⁄ √

( ) ( ) √ ( ) ( ) √

-7.9409 12.5409 = (-7.9409 , 12.5409)

Conclusion: This implies at 5% level of significance we have enough evidence / sufficient
evidence to conclude that the mean of new filtering device before installation and after
installation is the same.

Or we have 95% confident that that the difference mean of new filtering device before
installation and after installation is lie between -7.9409 and 12.5409.

[email protected]
21
Lecture notes for Design and analysis of experiments (Stat 2043) Chapter - 2

Case Four: If the two population variance is assumed to be unequal (  12   22 ), and unknown

and n1 and n2 are small (( )), then we estimate  12 and  22 by S12 and S 22 . In
this case: The test statistics becomes: -
2
 S12 S 22 
  
(̅ ̅ ) ( )  n1 n2 
( ) , Where U   2 2
 S12   S 22 
√
   
 n1    n2 
n1  1 n2  1
Then one can formulate the hypothesis as follows

Hypothesis Test Statistics Criteria for Rejection

( )
Reject H0 if | | ⁄

(̅ ̅ ) ( ) ( )
Reject H0 if
√ ( )
Reject H0 if

 A 100(1 − )% confidence interval for the difference in mean for the two independent
samples if is :-

{( ̅ ̅ ) ( )√ (̅ ̅ ) ( )√ }

Example: An experimenter was an interested in dieting and weight loss among men and women.
It was believed that in the first two weeks of standard dieting program, men would tend to loss
less weight than women. As check on this notion, a random sample of 15 women’s and 15 men
were put on the same diet and their mean weight losses are ̅ pound for men and
̅ pounds for women with variances of 2.56 and 0.31 respectively.
A, Did men weight losses significantly less than women at 5% level of significance?

Assume that the two populations are normally distributed with unequal variances

( ).

B, Make the 95% confidence interval in the mean difference of weight losses.

[email protected]
22
Lecture notes for Design and analysis of experiments (Stat 2043) Chapter - 2

Solution:

Given
n1  n2  15 ,

̅ , ̅
S12  2.56 , S 22  0.31

A,
Step1: 1   2  0 versus H 1 : 1   2  0
Step2:   0.05
Step 3: Identify the test statistic
t – Test, because the population variance is assumed to be unequal and small sample size.
Step 4: Determine the test Statistic
(̅ ̅ ) ( ) ( ) ( )

√ √

Step 5: Identify the critical value , t (U ) where

2
 S12 S 22   2.56 0.31 
2
     
U   n1 n2 
  15 15 
 17.34  17
2 2 2 2
 S12   S 22   2.56   0.31 
       
 n1    n2   15    15 
n1  1 n2  1 15  1 15  1

 t 0.05 (17 )  1.74

Step 6: Make decision, Reject H 0 if t cal  t (U )

Since -1.21 > -1.74, so we do not reject H 0 .

Step 7: Conclusion: at 5% level of significance we have enough evidence to say Men and
Women loss equal weight (women do not loss less weight than men).

[email protected]
23
Lecture notes for Design and analysis of experiments (Stat 2043) Chapter - 2

B, The 95% confidence is

(̅ ̅ ) ( )√ (̅ ̅ ) ( )√

2
 S12 S 22   2.56 0.31 
2
     
Where, U    n1 n2 
  15 15 
 17
2 2 2 2
 S12   S 22   2.56   0.31 
       
 n1    n2   15    15 
n1  1 n2  1 15  1 15  1

( ) √ ( ) √

-1.4129 = (-1.4129, 0.4329)

Conclusion: This implies at 5% level of significance we have enough evidence / Sufficient
evidence to conclude men and women loss equal weight (women do not loss less weight than
men) .
Or we have 95% confident that the mean of women weight loss and men weight loss is lie
between -1.4129 and 0.4329.

2.1.2 paired sample t - test

In this section more emphasis is given to selecting similar units. Select similar units, form pairs
and then apply the treatments or apply the same treatment on the same units but before and after
some condition.
The aim of paring is to make more accurate comparison by having members in pair as like as
possible except difference due to treatments that the investigator deliberately introduces.
o In paired sample the sample size in each treatment must be equal but the sample size is
small (i.e. n ).
The actual analysis of paired data requires us to compute the differences in the n pairs of
measurements, = , and, obtain ̅ , SD, the mean and standard deviations in the
Dis. Also, we must formulate the hypotheses about the mean of the differences, µD = 0.

[email protected]
24
Lecture notes for Design and analysis of experiments (Stat 2043) Chapter - 2

The conditions required developing at procedure for testing hypotheses and constructing
confidence intervals for µD are:
The test Statistics is as follows:-
̅ ∑ ∑( ̅) ∑ ̅
tcal = , where ̅ √
√

Then one can formulate the hypothesis as follows

Hypothesis Test Statistics Criteria for Rejection
Reject H0 if | |>
̅
tcal = Reject H0 if > tα , n-1
√

Reject H0 if tcal < -tα , n-1

 A 100(1-α) % confidence interval for the difference in mean for the paired samples is:

̅ - tα/2 n-1 ̅ + tα/2 n-1 Or

√ √

̅ ± tα/2 n-1 , where n is the number of pairs of observation

√

Example: The following are physiological measures taken of patients before and after the
administration of some medication. Assume that the data are from normally distributed
populations.
Patient 1 2 3 4 5 6 7 8 9 10
Before 101 96 98 102 95 99 107 103 102 104
After 111 110 107 101 121 115 122 118 123 105
Di -10 -14 -9 1 -26 -16 -15 -15 -21 -1

a) Test for significant difference between the mean measures before and after the
medication at 5% level of significance.
b) Construct a 95% confidence interval for the true mean change.

[email protected]
25
Lecture notes for Design and analysis of experiments (Stat 2043) Chapter - 2

Solution:
n

D i
 126
DD i 1
  12.6
n 10

 D  D
n
2
i
 S D2  i 1
 68.27  S D  68.27  8.26
n 1
S D2 S 8.26
S .E ( D )  Var( D )   D   2.61
n n 10
H0 : D  0
a), Step 1: versus H 1 :  D  0
Step 2:   0.05
Step 3: Identify the test statistic.
The test statistic is t – test because the sample size is small.
Step 4: Determine the test statistic.
D  D  12.6  0
=   4.83
S D2 n 2.61

Step 5: Identify the critical value,

The critical value is t 2 (n  1)  t 0.05 2 (10  1)  t 0.025 (9)  2.262

Step 6: Make Decision, Reject H 0 if t cal  t 2 (n  1)

Since  4.83  2.262 , so we reject H 0

Step 7: Conclusion:- at 5% level of significance there is strong evidence to conclude that

there is significant difference in mean physiological measure before and after the
medication.
b), A 100(1-α) % CI for is:
t 2 (n  1)  t 0.05 2 (10  1)  t 0.025 (9)  2.262

̅ - tα/2 n-1 ̅ + tα/2 n-1 = -12.6 – 2.262 -12.6 + 2.262

√ √ √ √

-18.5084 -6.6916 = (-18.5084, - 6.6916)

We have 99% confident that the mean difference of physiological measure before and after
medication is lie between -18.5084 and -6.6916).

[email protected]
26
Lecture notes for Design and analysis of experiments (Stat 2043) Chapter - 2

2.2 Inference about the Variance of Normal Distribution

A, Hypothesis Testing and Interval Estimation for population Variance.

 The sample variance is given by:-

∑ ( ̅)

To obtain sampling distribution of S 2 , sample should be taken from normally distributed

population with mean  and variance  2 .

 Note: To test whether the population variance  2 , is equal to some value  o2 or not, we
use the sampling distribution of sample variance. If sample size n is taken from normally
distributed population , then
( )
, Then one can formulate the hypothesis as follows

Hypothesis Test Statistics Criteria for Rejection

Reject H0 if > ,(n-1)
Or Reject H0 if < ,(n-1)

Reject H0 if ( )
( )

Reject H0 if ( )

 A 100 1   % Confidence Interval (CI) for true value of population variance  2 is:

( ) ( )
( ) ( )

( ) ( )
Or ( )
( ) ( )

Note: If you get once the CI for  2 , then a 100 1   % Confidence Interval (CI) for  can be

obtained by taking a positive square root of confidence limits to  2 .

[email protected]
27
Lecture notes for Design and analysis of experiments (Stat 2043) Chapter - 2

Example 1: Based on past data, the variance of a normal population is hypothesized to be 48. If
a sample of 15 observations yields a variance of 56.

A, Test the hypothesis that the variance has increased? Use 1% level of significance.
B, Construct 99% confidence interval for population variance  2 .

Solution: Given:-  02  48, n  15 and S 2  56

A,

H 0 :  2  48
.
Step 1:
H 1 :  2  48

Step 2:   0.01
Step 3: Identify the test statistic. The test statistic is .

(n  1) S 2 14 * 56
Step 4: Determine the test statistic,  cal
2
   16.33
 2
0 48

Step 5: Identify the critical value, the critical value is  2 (n  1)   02.01 (14 )  29 .141

Step 6: Decision, Reject H 0 if  cal

2
  2 (n  1)

Since 16.33 < 29.141, so don’t reject H 0

Step 7: Conclusion, at 1% level of significance we have no enough evidence to reject H 0 .

Therefore, the variance of the population has not increased.
B,
The 99% confidence interval is
( ) ( )

   
 n  1S 2 n  1S 2   15  156 15  156 
 2 , 2  =  2 , 2 
   , (n  1) 1 , (n  1)    0.01 , (15  1) 1 0.01 , (15  1) 
 2 2   2 2 

 15  156 15  156 

=  ,  = (25.0327 , 192.3926 )
 31.319 4.075 
Conclusion: we have 99% confident that the population variance is lie between 25.0327 and
192. 3926.

[email protected]
28
Lecture notes for Design and analysis of experiments (Stat 2043) Chapter - 2

Example 2: A company manufacturing radio tubes for the last 10 years found that the life of
their tubes has variance of 0.6 year2. As a result of some qualitative improvement, the company
claims that the variance of life of their tubes has decreased. If the sample variance of 9 randomly
selected tubes is found to be 0.45 year2
A, Using 0.05 level of significance, test the claim made by the company.
B, Make the 95% confidence interval for population variance, .
Solution: Given:  02  0.6, n  9 and S 2  0.45

Step 1: versus

Step 2:   0.05

Step 3: the test statistic is

(n  1) S 2 8 * 0.45
Step 4: Determine the test statistics, the test Statistic is  2
  6

cal 2
0 0.6

Step 5: Identify the critical value, the critical value is 12 (n  1)  120.05 (8)   02.95 (8)  2.733

Step 6: Make decision, Reject H 0 if  cal

2
  12 (n  1)

Since 6 > 2.733, so don’t reject H 0

Step 7: Conclusion: at 5% level of significance we have enough evidence to accept H 0 ,

Therefore, the variances of life time of tubes has no reduced / decreased.

B, The 95% confidence interval is
The 99% confidence interval is
( ) ( )

   
 n  1S 2 n  1S 2   9  10.45 9  10.45 
 2 , 2 =  2 , 2 
   , (n  1) 1 , (n  1)    0.05 , (9  1) 1 0.05 , (9  1) 
 2 2   2 2 

 9  10.45 9  10.45 
=  ,  = (0.2053, 1.6514)
 17.535 2.180 
Conclusion: we have 95% confident that the population variance is lie between 0.2053 and
1.6514.

[email protected]
29
Lecture notes for Design and analysis of experiments (Stat 2043) Chapter - 2

B. Hypothesis Testing and Interval Estimation for Comparing Two Population Variances

 In many situations we need to compare the standard deviations of the populations.

 Another major application of a test for the equality of two population variances is for

evaluating the validity of the equal variance condition (that is,  12 =  22 )

 In order to compare two population variance the appropriate test statistics is :-

A statistical test comparing  1 and  2 utilizes the test statistic s12 / s22 . When  12 =
2 2

 22 ,  12 /  22 =1 and s12 / s22 follows an F distribution with df1 = n1  1 and df 2 = n2  1

Then one can formulate the hypothesis is as follows

Hypothesis Test Statistics Criteria for Rejection

Reject H0 if > F (n1  1, n2  1)
2 Or
Reject H0 if <F  ( n1  1, n 2  1)
1
2

Reject H0 if Fcal  F (n1  1, n2  1)

Reject H0 if > ( )

NB: ( ) ( )

[email protected]
30
Lecture notes for Design and analysis of experiments (Stat 2043) Chapter - 2

 General Confidence Interval for  12 /  22 with Confidence level 1   

 When comparing population variances,  12 /  22 the appropriate measure is the ratio of the population

variances: Thus, we need to formulate a confidence interval for the ratio  12 /  22 ,

 A 100 1    % confidence interval for the ratio the two population variance (  12 /  22 ) is given by.

( ) ( )

( ( ) ( ))

Note: If you get once the CI for , then a 100 1   % Confidence Interval (CI) for the ratio of

population standard deviation can be obtained by taking a positive square root of confidence

limits to .

Example: Suppose that independent random samples, one consisting 13 cases and the other
consisting 9 cases were drawn from two normal populations. Sample standard deviations are
found to be 48.1 and 89.2 respectively.
A, Test whether the two populations have equal variances at 5% level of significance.

B, Make 95% confidence interval for the ratio of the two population variance (  1 /  2 ).
2 2

Solution: Given, S1  48 .1, S 2  89 .2, n1  13, n2  9

A,

 12
. H0 : 1
 22
Step 1:
 12
H1 : 2  1
2
Step 2:   0.05
Step 3: Identify the test statistic:

[email protected]
31
Lecture notes for Design and analysis of experiments (Stat 2043) Chapter - 2

S12 48.12
Step 4: Determine the test statistic, Fcal  2   0.29
S 2 89.2 2
Step 5: Identify the critical value, the critical value is
F (n1  1, n2  1)  F0.025 (12, 8)  4.20
2

Step 6: Make decision, Reject H 0 if Fcal  F (n1  1, n2  1)

2

, so don’t reject .
Step 7: Conclusion
 Therefore, at 5% level of significance we have enough evidence to conclude that , there is no
significance difference between the two variances (the two variances are equal). Or we have
found that insufficient statistical evidence to conclude that the variance of the two population
variance is different.

B, The 95% confidence interval for the ratio of is :

( ) ( )

( ) ( ) ( )
,
( ) ( ) ( ) ( )

( )

= 0.0692 = (0.0692, 1.0206)

Conclusion: we have 95% confident that the ratio of population variance ( ) is lie between

0.0692 and 1.0206.

[email protected]
32
Lecture notes for Design and analysis of experiments (Stat 2043) Chapter - 2

Exercise: A chemical engineer is investigating the inherent variability of two types of test
equipment that can be used to monitor the output of a production process. He suspects that the old
equipment , type 1 , has a larger variance than the new one .

A, At 5% level of significance, test the engineer claim?

B, Make 95% confidence interval for the ratio of the two population variance (  1 /  2 ).
2 2

[email protected]
33