Advertising 3

SYLLABUS
1 of 71
http://www.vet.pctiltd.com/PCTILMS/courses/BBA 4th Semester/BBA-0...
Not for sale

For personal use of KSOU
Information technology & Management programme
student only
KARNATAKA STATE OPEN UNIVERSITY

MANASAGANGOTRI, MYSORE-570006
BBA-041
ADVERTISING FUNDAMENTAL
BLOCK 4 APPLICATION IN RESEARCH METHODOLOGY
In collaboration w ith
VIRTUAL EDUCATION TRUST

Registered Office
UU -11, Pitampura, Delhi 110 034
Ph : 011-27348396/47510411
Fax : 011-27347796 / 47591499
E : vet@pctiltd.com, W : http://www.vet.pctiltd.com
8/19/2012 12:08 AM
SYLLABUS
2 of 71
First Edition 2009
Copyright by Virtual Education Trust

All right reserved
No part of this publication may be reproduced

in any form or by any means without the written
permission from the Virtual Education Trust
All product names and company names mentioned

herein are the property of their respective owner.
8/19/2012 12:08 AM
SYLLABUS
3 of 71
Block 4 Application in Research Methodology

Unit 1- Testing Of Hypothesis-Large and Small Samples: Objective, Introduction, Testing of
Hypothesis for Proportions, Hypothesis tests of Differences between Means, Theoretical Basis,
Standard Error of the Difference between Two Proportions, Uses of p- Values, Using the
computer to Test the Hypotheses, T Test for Difference between Two Sample Means,
Measuring the power of a Test, Theoretical aspects of the t distribution, Characteristics of the t
distribution, Degree of Freedom, t Values for one Tailed Tests, Hypothesis Testing Using the t
Distribution, Two Tailed Test, Tests for Differences between Means Small Samples, Degrees of
Freedom.
Unit 2- Non Parametric Tests: Objective, Introduction, Non Parametric Tests, Important Types of
Nonparametric Tests, Non-parametric vs. Distribution-free Tests, Standard uses of Non
Parametric Tests, Tests for Ordinal Data, The Null Hypothesis.
Unit 3- Chi-Square Test: Objective, Introduction, Key Definitions, Properties of the Chi-Square,
Chi-Square Probabilities, Degree of Freedom Which Arent in the Table, Uses of Chi-square
Test, Applications of Chi-square Test.
Unit 4- Analysis of Variance (ANOVA): Objectives, Introduction, Logic behind ANova, Applications of
Anova, Dependent and Independent Variables.
8/19/2012 12:08 AM
SYLLABUS
4 of 71
Advertising Fundamental
Application in Research Methodology
Unit 1
Testing Of Hypothesis-Large and Small Samples 2
Unit 2
Non Parametric Tests 34
Unit 3
Chi-Square Test 53
Unit 4
Analysis of Variance (ANOVA) 73
8/19/2012 12:08 AM
SYLLABUS
5 of 71
Unit 1 Testing Of Hypothesis-Large and Small Samples

Structure
1.0 Objectives
1.1 Introduction
1.2 Testing of Hypothesis for Proportions
1.3 Hypothesis tests of Differences between Means
1.3.1 Theoretical Basis
1.4 Standard Error of the Difference between Two Proportions
1.5 Uses of p- Values
1.6 Using the computer to Test the Hypotheses
1.7 T Test for Difference between Two Sample Means
1.8 Measuring the power of a Test
1.9 Theoretical aspects of the t distribution
1.10 Characteristics of the t distribution
1.11 Degree of Freedom
1.12 t Values for one Tailed Tests
1.13 Hypothesis Testing Using the t Distribution
1.14 Two Tailed Test
1.15 Tests for Differences between Means Small Samples
1.15.1 Degrees of Freedom
1.16 Summary
1.17 Questions
1.18 References
8/19/2012 12:08 AM
SYLLABUS
6 of 71
1.0 Objective
After studying this unit you will be able to:
Describe testing of hypothesis
Discuss sample proportions
Explain test for the differences between two samples.
1.1 Introduction
Let us understand about the principles of hypothesis testing where we have compared a sample
mean with a hypothesized or population mean.
We can also apply the principles of hypothesis testing already determined to testing for
differences between proportions of occurrence in a sample with a hypothesized level of
occurrence.
Theoretically the binomial distribution is the correct distribution to use when dealing with
proportions. As sample size increases the binomial distribution approaches a normal distribution
in terms of characteristics. Therefore we can use the normal distribution to approximate the
sampling distribution. To do this we need to satisfy the following conditions:
np>5, nq>5
where p is the proportion of successes
q: proportion of failures.
1.2 Testing of Hypothesis for Proportions

When testing significance of hypotheses for proportions we begin with a similar procedure to the
earlier case:First we define various proportions or percentages of occurrence of a event:
p Ho: the population or hypothesized population of success
q Ho Hypothesized value of the population proportion of failures
p: sample proportion of successes
q: sample proportion of failures
Again we calculate a z statistic
where p is the standard error of the proportion which is calculated using the hypothesized
population proportion.
8/19/2012 12:08 AM
SYLLABUS
7 of 71
A ketchup manufacturer is in the process of deciding whether to produce a new extra spicy
ketchup. The companys marketing research department used a national telephone survey of
6000 households and found that 335 would purchase extra spicy ketchup. A much more
extensive study made two years ago found showed that 5% of households would purchase would
purchase the brand then. At a two percent level of significance should the company conclude
that there is an increased interest in the extra spicy flavour?
z critical for a one tailed test is 2.05 Since the observed z > z critical we should reject.Ho and
current levels of interest are significantly greater than interest two years ago.
1.3 Hypothesis tests of Differences between Means

So far we have examined the case where we are testing the results of a sample against a
hypothesized value of a population statistic.
We now turn to case where we wish to compare the parameters for two different populations and
determine whether these differ from each other. In this case we are not really interested in the
actual value of the two parameters but the relation between the two parameter, i.e. is there a
significant difference between them.
Example of hypothesis of this type are:
Whether female employees earn less than males for the same work.
A drug manufacturer may need to compare reactions of one group of animals administered
the drug and the control group.
A company may want to see if the proportion of promotable employees in one installation is
different from another.
In each case we are not interested in the specific values of the individual parameters as the
relation between the two parameters.
The core problem reduces to one of determining whether the means from two samples taken from
two different populations is significantly different from each other. If they are not, we can
hypothesize that the two samples are not significantly different from each other.
8/19/2012 12:08 AM
SYLLABUS
8 of 71
1.3.1 Theoretical Basis

Shown below are three different but related distributions.
Figure 1a shows the population distribution for two different populations 1 and 2.
They have respectively the following characteristics:
Mean and and standard deviations s and s
1 2 12.
Figure 1b shows the respective sampling distribution of sample. This distribution is defined by the
following statistics:
Mean of sampling distribution of sample means : 1 and 2
Standard deviation of sampling distribution of mean or the standard error of sampling mean: s
and s2.
x1 x
We have two populations 1, 2 with mean 1 and 2 with standard deviation s and s The associated
sampling distribution for 12.
sampling means x1 and x2 However what we are now interested in is the difference between the
two values of the sampling means. i.e., the sampling distribution of the difference between the
sampling means x1 - x2. How do we derive this distribution ?
Suppose we take a random sample from the distribution of Population 1 and another random
sample from the distribution of Population 2. If we then subtract the two sample means, we get x1
- x2 . This difference will be positive if x1 is larger than x2 and vice versa. By constructing a
distribution of all possible sample differences x1 - x2 we end up with a distribution of the
difference between sample means shown below in figure 1.
8/19/2012 12:08 AM
SYLLABUS
9 of 71
The mean of this distribution is x x = 1 1-22

The standard deviation of the distribution of the difference between sample means is called the
standard error of the difference between two means.
The testing procedure for a hypothesis is similar to the earlier cases.
Since we will usually we will be testing for equality between the two population means hence:
(1 -) o =0 since 1 = 2
2H
An example will make the process clearer:
A manpower statistician is asked to determine whether hourly wages of semi skilled labour are
the same in two cities. The result of the survey is given in the table below. The company wants to
test the hypothesis at .05 level of significance that there is no significant difference between the
hourly wage rate across the two cities.
8/19/2012 12:08 AM
SYLLABUS
10 of 71
Since standard deviation of the two populations are not known we estimate s^ s^ by using the
sample standard deviation s and
The estimated standard error of the difference between the two means is
We then calculate the z statistic:
We can mark the standardized difference on a sketch of the sampling distribution and compare
with the critical value of z=1.96 in figure 7 As we can see the calculated z lies outside the
acceptance region. Therefore we reject the null hypothesis.
Example 2
1. Two independent samples of observations were collected. For the first sample of 60 elements,
the mean was 86 and the standard deviation 6. The second sample of 75 elements had a mean
of 82 and a standard deviation of 9.
a. Compute the estimated standard error of the difference between the two means.
b. Using a = 0.01, test whether the two samples can reasonably be considered to have come from
populations with the same mean.
8/19/2012 12:08 AM
SYLLABUS
11 of 71
They limits of the acceptance region are : 2.58
Since 3.09>2.58, we reject Ho and it is reasonable to conclude that the two samples come from
different populations.
1.4 Standard Error of the Difference between Two Proportions

Since we do not know the population proportions, we need to estimate them from sample
statistics: p 1 , p2 , q 1 , q1
We hypothesize that there is no difference between the two proportions. In which case our best
estimate of the overall population proportion successes is the combined proportion of successes
in both samples. This can be calculated from the following formula:
p=p 1 q 1+ p 2 q 2 / n1+ n 2
Once we have the estimated proportion of successes in two populations. The estimated standard
error of the difference between two proportions is as follows:
The standard z statistic in this case is calculated as :
This can be best illustrated with the help of an example:

A pharmaceutical company tests two new compounds to reduce blood pressure. They are
administered to two different groups of lab animals. In group1 71 of 100 animals respond to
drug1 with lower BPS. In group2 58 of 90 animals showed lower BP levels.
The company wants to test at .05 level of significance whether there is a difference between the
two drugs.
8/19/2012 12:08 AM
SYLLABUS
12 of 71
The overall population or hypothesized percentage of occurrence assumes that there is no

difference between the two population proportions. We therefore estimate it as the combined
proportion of successes in both samples.
The estimated standard error of the difference between two sample proportions is:
The observed z statistic is:
z. critical for .o5 level of significance is 1.96. Since observed z is less than z critical we accept Ho.
This is shown in the figure 3 below:
Example 3:
For tax purposes a city government requires two methods of listing property. One requires
property owner to appear in person before a tax lister. The second one allows the form to be
mailed. The manager thinks the personal appearance method leads to lower fewer mistakes. She
authorizes an examination of 50 personal appearances, and 75 mail forms. The results show that
10% of personal appearances have errors whereas 13.3 % of mails forms had errors. The
manager wants to test at the .15 level of significance, the hypothesis that that personal
appearance method produces lower errors. The hypothesis is a one tailed test. The procedure for
this as the same as for carrying out a one tailed test for comparing sample means. The data is as
follows:
p 1: .1 q 1: .9 n 1 : 50
8/19/2012 12:08 AM
SYLLABUS
13 of 71
p 2: ..133 q 12: ..867 n2 : 75

Ho: p1= p2 Ha: p 1< p2 a=.15
Since it is a one tail test we do not divide the level of significance on both sides of the normal
curve. A .15 level of significance implies we determine z critical for area under one side of the
normal curve i.e., .35 (.5-.15). Specifically it is a left tailed test as the manager wishes to test that
method 1 , i.e., personal appearances, result in significantly lower errors. This is shown by the
marked of region in the figure4
therefore calculated z <critical z we accept the null hypotheses.

Hint: If a test is concerned with whether one proportion is significantly different from another, use
a two-tailed test. If the test asks whether one proportion is significantly higher or lower than the
other, a one tailed test is appropriate.
1.5 Uses of p- Values

Use of p values saves the tedium of looking up tables. The smaller the probability value, the
greater the significance of the finding. The simple rule of thumb is: As long as a>p reject Ho.
For example if we have a p-value=.01 and =.05 . Then this means that the probability of getting
our sample result is .01. We compare this with our standard of accepting/ rejecting Ho which is
.05. since .05>.01 we reject Ho as the probability of getting such a result is much lower than our
level of significance.
Example 4
The Coffee Institute has claimed that more than 40% of American adults regularly have a cup of
coffee with breakfast. A random sample of 450 individuals showed that 200 of them were regular
coffee drinkers at breakfast. What is the prob value of a test of hypotheses seeking to show that
the Coffee Institutes claim was correct.
8/19/2012 12:08 AM
SYLLABUS
14 of 71
1.6 Using the computer to Test the Hypotheses

These days in actual managerial situations hypotheses tests are rarely done manually. Therefore
it is important that students can interpret computer output generated for hypotheses tests by
various standard statistical analysis packages. The most popular of these packages are SPSS
and Minitab. Broadly all programmes follow the same principles. Instead of a comparing the
calculated z value with a predetermined level of significance, most packages display the prob
values or p- values.
To accept or reject an hypotheses we compare the level of significance (a) and the p- value. If
a>p- value we reject Ho at the relevant level of significance and vice versa.
An example will help show how computer outputs results for hypotheses testing.
Example 5:
While designing a test it was expected that the average grade would be 75%, i.e., 56.25 out of 75.
This hypotheses was tested against actual test results for a sample of 199 students.
This hypotheses was tested at a =.05

The computer output for this test is shown below using the Minitab package. The observed t
value for this test was -15.45, with an associated (two-tailed) prob value of 0.0000. Because this
prob value is less than our significance level of = 005 we must reject Ho and conclude that the
test did not achieve the desired level of difficulty.
This is shown in Table 1
1.7 T Test for Difference between Two Sample Means

Here we test hypotheses of equality of two means.The university had been receiving many
complaints about the caliber of teaching being done by the graduate-student teaching assistants.
As a result, they decided to test whether students in sections taught by the graduate TAs really
did worse in the exam than those students in sections taught by the faculty.
If we let the TAs sections be sample 1 and the facultys sections be sample 2, then the
appropriate hypotheses for testing this concern are
8/19/2012 12:08 AM
SYLLABUS
15 of 71
The underlying population is assumed to be equal for both samples.

The Mnitab output for doing this test is given below. The test results are reported assuming that
the two population variances are equal. If we can assume that the two variances are equal, then
the test reported by Minitab is the test using a pooled estimate for S z This is shown in table 2
Both use pooled stdev=10.0

What does the data tell us regarding the efficacy of TA ? The prob value is quite high, i.e., .33.
Therefore if we compare this with a level of significance of .05 (a = 0.05) we would accept the null
hypotheses that there is no difference in results between TAs and faculty. In this case we would
accept the null hypotheses at any level of significance up to .33.
1.8 Measuring the power of a Test

What should a good hypothesis test do ?
Ideally and (the probabilities of Type I and Type 2 errors) should both be small. A Type I error
occurs when we reject a null hypothesis that is true. a (the significance level of the test) is the
probability of making a Type I error. Once we decide on the significance level, there is nothing
else we can do about a.
A Type 2 error occurs when we accept a null hypothesis that is false; the probability of a Type 2
error is . Ideally a manager would want a hypothesis to reject a null hypothesis when it is false.
Suppose the null hypothesis is false. Then managers would like the hypothesis test to reject it all
the time. Unfortunately, hypothesis tests cannot be foolproof; sometimes when the null
hypothesis is false, a test does not reject it, and a Type 2 error is made.
When the null hypothesis is false, (the true population mean) does not equal Ho (the
hypothesized population mean); instead, equals some other value. For each possible value of for
which the alternative hypothesis is true, there is a different probability of incorrectly accepting the
null hypothesis.
Of course, we would like this (the probability of accepting a null hypothesis when it is false) to be
as small as possible or we would like 1- (the probability of rejecting a null hypothesis when it is
8/19/2012 12:08 AM
SYLLABUS
16 of 71
false) to be as large as possible.

Therefore a high value of 1 - (something near 1.0) means the test is working quite well ( i.e it is
rejecting the null hypothesis when it is false); a low value of 1 - (something near 0.0) means that
the test is working very poorly ( i.e. not rejecting the null hypotheses when it is false).
The value of 1 - is a measure of how well the test is working and is known as the power of the
test. If we plot the values of 1 - for each value of for which the alternative hypothesis is true, the
resulting curve is known as a power curve. This can be explained better with the help of an
example.
Example
We were deciding whether to accept a drug shipment. Our test indicates that we should reject
the null hypothesis if the standardized sample mean is less than - 1.28, that is, if sample mean
dosage is less than 100.00 - 1.28 (0.2829), or 99.64 cc.
In Figure, we show a left-tailed test. In Figure 9b, we show the power curve which is plotted by
computing the values of 1 - .
Point C on the power curve in Figure 2 b shows population mean dosage is 99.42 cc. Given that
8/19/2012 12:08 AM
SYLLABUS
17 of 71
the population mean is 99.42 cc, we must compute the probability that the mean of a random
sample of 50 doses from this population will be less than 99.64 cc (the point below which we
decided to reject the null hypothesis i.e., the value of the dose at which we rejected the null
hypothesis. This shown in Figure 2c .
We had computed the standard error of the mean to be 0.2829 cc. So 99.64 cc is (99.6499.42)/0.2829 = 0.78
Thus 99.64 is .78 SE above the true population mean when it takes a value = 99.42 cc.
The probability of observing a sample mean less than 99.64 cc and thus rejecting the null
hypothesis is 0.7823, when we take the true population mean to be = 99.42 cc. This is given by
the colored area in Figure 9c. Thus, the power of the test 1 - at = 99.42 is 0.7823. This simply
means that if = 99.42, the probability that this test will reject the null hypothesis when it is false is
0.7823.
Point D in Figure 9b shows that if the population mean dosage is 99.61 cc. We then ask what is
the probability that the mean of a random sample of 50 doses from this population will be less
than 99.64cc and thus cause the test to reject the null hypothesis?
This is illustrated in Figure 2d. Here we see that 99.64 is (99.64 -99.61)/0.2829, or 0.11 standard
error above 99.61 cc. The probability of observing a sample mean less than 99.64cc and thus
rejecting the null hypothesis is 0.5438, the colored area in Figure 9d. Thus, the power of the test
(1 - ) at = 99.61 cc is 0.5438.
Using the same procedure at point E, we find the power of the test at = 99.80 cc is 0.2843; this is
8/19/2012 12:08 AM
SYLLABUS
18 of 71
illustrated as the colored area in Figure 2e.
As we can see the values of 1 - continue to decrease to the right of point E. This is because as
the population mean gets closer and closer to 100.00 cc, the power of the test (1 - ) gets closer
and closer to the probability of rejecting the null hypothesis when the population mean is exactly
100.00 cc. This probability is nothing but the significance level of the test which in this case is
0.10. The curve terminates at point F, which lies at a height of 0.10 directly over the population
mean.
What does the power curve in Figure 2b tell us?
As the shipment becomes less satisfactory (as the doses in the shipment become smaller), our
test is more powerful (it has a greater probability of recognizing that the shipment is
unsatisfactory). It also shows us, however, that because of sampling error, when the dosage is
only slightly less than 100.00 cc, the power of the test to recognize this situation is quite low.
Thus, if having any dosage below 100.00 cc is completely unsatisfactory, the test we have been
discussing would not be appropriate.
1.9 Theoretical aspects of the t distribution

Theoretical work on the t distribution was done by W.S. Gosset in the 1900s.
The students t distribution is used under two circumstances:
Sample size, n , is less than 30.
Where population standard deviation is not known. In this case t tests may be used even if the
sample size is greater than 30.
We also assume that the population underlying a t distribution is normal or approximately normal.
1.10 Characteristics of the t distribution

Relationship between the t distribution and normal distribution:
1. Both distributions are symmetrical. However as can be seen in figure1 the t distribution is flatter
than the normal distribution and is higher in the tails and has proportionately less area in the
around the mean. This implies that we have to go further out from the mean of a t distribution
8/19/2012 12:08 AM
SYLLABUS
19 of 71
to include the same area under the curve. Thus interval widths are much wider for a t
distribution.
2. There is a different t distribution for every possible sample size.
3. As sample size increases, the shape of the t distribution loses its flatness and becomes
approximately equal to the normal distribution. In fact for sample sizes greater than 30 the t
distribution becomes less dispersed and approximates a normal distribution and we can use
the normal distribution.
1.11 Degree of Freedom

What is Degree of Freedom?
This is defined as the number of values we can choose freely. The concept is best illustrated with
the help of an example:
Consider the case: a+b/2=18
Given that the mean of these two numbers has to equal 18, how do we determine values for a
and b? Basically we can slot in any two values such that they add up to 36. Suppose a=10. then
b has to equal 26 given the above constraint. Thus in a sample of two where the value of the
mean is specified ( i.e., a constraint) we are only free to specify one variable. Therefore we have
only one degree of freedom.
Another example: a+b+c+d+e+f+g/7=16
Now we have 7 variables. Given the mean we are free to specify 6 variables. The value of the 7 th
variable is determined automatically.
For a sample size of n we can define a t distribution for degree of freedom n-1.
Using The t Distribution Tables
The t table differs in construction from the normal table in that it is more compact. It shows
areas under the curve and t values for a limited number of level of significance (usually .01,
.05, 10). t values are therefore defined for level of significance and degrees of freedom.
A second difference is that we must specify the degrees of freedom with which we are dealing.
Suppose we are making an estimate for a n=14, at 90% level of confidence. We would go
down vertically to determine the degrees of freedom (i.e. 13) and then read of the appropriate
t value for a level of significance of .1.
8/19/2012 12:08 AM
SYLLABUS
20 of 71
The normal tables focus on the chance of that the sample statistic lies within a given number
of standard deviations on either side of the population mean. The t distribution tables on the
other hand measures the chance that the observed sample statistic will lie outside it our
confidence interval, defined by a given number of standard deviations on either side of the
mean. A t value of 1.771 shows that if we mark off plus and minus 1.771s x = on either side of
the mean then we enclose 90% of the area under the curve. The area outside these limits, i.e.,
that of chance error, will be 10%.This is shown in the
Figure 2 below. Thus if we are making an estimate at the 90% confidence limit we would look in
the t tables under the .1 column (1.0-.9=.1). This is actually or the probability of error
1.12 t Values for one Tailed Tests

The procedure for using t tests for a one tailed test is conceptually the same as for a one tailed
normal test. However the t tables usually give the area in both tails combined at a specific level of
significance.
For a one tailed test t test, we need to determine the area located in only one tail. For example to
find the appropriate t value for a one tailed test at a level of significance of .05 with 12 degrees of
freedom we look in the table under the .10 column opposite 12 degrees of freedom. The t value is
1.782. This is because the .10 column represents .10 of the area contained under both tails
combined. Therefore it also represents .05 of the area contained in each tail separately.
Exercise
Find one tail value for n=13, a=.05 % degrees of freedom=12
T value for one tail test we need to look up the value under the .10 column t= 1.782
Find one tail t values for the following:
n=10, a=.01
n=15, a=.05
1.13 Hypothesis Testing Using the t Distribution

The procedure for hypothesis testing using the t test is very similar to that followed for the normal
test. Instead of calculating the z statistic we calculate a t statistic. The formula for the t statistic is
8/19/2012 12:08 AM
SYLLABUS
21 of 71
where
is the estimated standard error of the sample means.
The t test is the appropriate test to use when population standard deviation is not known and has
to be estimated by the sample standard deviation.
where s is the sample standard deviation
This represents the basic t test. Variants of this formula are developed to meet the requirements
of different testing situations. We shall look at more common types of problems briefly. As the
theoretical basis of hypothesis is the same as the normal distribution and has been dealt with in
detail in the last chapter, we shall focus on applications of the t test to various situations.
1. Hypotheses testing of means
The t test is used when :
1. the sample size is <30
or
2. When population standard deviation not known and has to be estimated by the sample
standard deviation.
3. When a population is finite and the sample accounts for more than 5% of the population we
use the finite population multiplier and the formula for the standard is modified to;
Exercise
Find one tail value for n=13, a=.05 % degrees of freedom=12
T value for one tail test we need to look up the value under the .10 column t= 1.782
Find one tail t values for the following:
n=10, a=.01
n=15, a=.05
1.14 Two Tailed Test:
8/19/2012 12:08 AM
SYLLABUS
22 of 71
The specification of the null and alternative hypotheses is similar to the normal distribution.
Ho: =
o
Ha:
o
This is tested at a prespecified level of significance The t statistic is
The calculated t value should be compared with the table t value. If t calculated< t critical we
accept the null hypotheses that there is no significant difference between the sample mean and
the hypothesized population mean. If the calculated t value > t critical we reject the null
hypotheses at the given level of significance.
An example shall make the process clearer:
A personnel specialist is a corporation is recruiting a large number of employees. For an overseas
assignment. She believes the aptitude scores are likely to be 90. a management review finds the
mean scores for 20 test results ot be 84 with a standard deviation of 11. Management wish to test
the hypotheses at the .10 level of significance that the average aptitude score is 90.
Our data is as follows;
As we can see this represents a two-tailed test.

Degrees of freedom=19
To find t critical we look under the t table under the .10 column, which gives the t value for .05
under both sides of the t curve. t. =1.729
As population standard deviation is not known we estimate it :
s
=11 where s is the sample standard deviation
Standard error of sampling mean
8/19/2012 12:08 AM
SYLLABUS
23 of 71
Therefore since -2.44< -1.729 we reject the personnel managers hypotheses that the true mean
of employees being tested is 90. This is also illustrated diagrammatically in figure 3
Exercises
Given a sample mean 83, Given a sample mean of 94.3, a sample standard deviation of 12.5
and a sample size of G size of 22, test the hypothesis that the value of the population mean is
70 against the alternative the hypothesis that it is more than 100. Use the 0.025 significance
level.
If a sample of 25 observations reveals a sample mean of 52 a sample variance of 4.2, test the
hypothesis that the population mean is 05 against the alternative hypothesis that it is some
other value. Use the .01 level of significance. .
Picosoft, Ltd., a supplier of operating system software for personal computers, was planning
the initial public offering of its stock in order to raise sufficient working capital to finance the
development of a new seventh-generation integrated system. With current earnings $1.61 a
share, Picosoft and its underwriters were contemplating an offering price of $21, or about 13
times earnings. In order to check the appropriateness of this price, they randomly chose seven
publicly traded software firms and found that their average price/ earnings ratio was 11.6, and
the sample standard deviation was 1.3. At = .02 can Picosoft conclude that the stocks of
publicly traded software firms have an average P /E ratio that is significantly different from 13?
The data-processing department at a large life insurance company has installed new color
video display terminals to replace the monochrome units it previously used. The 95 operators
trained to use the new machines averaged 7.2 hours before achieving a satisfactory level of
performance. Their sample variance was 16.2 squared hours. Long experience with operators
on the old monochrome terminals showed that they averaged 8.1 hours on the machines
before their performances were satisfactory. At the 0.01 significance level, should the
supervisor of the department conclude that the new terminals are easier to learn to operate?
1.15 Tests for Differences between Means Small Samples

Again broadly the procedure for testing whether the sample means from two different samples
8/19/2012 12:08 AM
SYLLABUS
24 of 71
are not significantly different from each other is the same as for the large sample case. The
differences are in the calculation of the standard error formula and secondly in the calculation of
the degrees of freedom.
1.15.1 Degrees of Freedom
In the earlier case where we had tested the sample against a hypothesized population value, we
had used a t distribution with n-1 degrees of freedom. In this case we have n 1 1 degrees of
freedom for sample 1 and n2 1 for sample 2. When we combine the sample to estimate the
pooled variance we have n 1 + n 2 2 degrees of freedom . Thus for example if n 1 =10 and n2 = 12
the combined degrees of freedom = 20
Estimation of Sample Standard Error of the difference Between Two Means.
In large samples had assumed the unknown population variances were equal and we estimated
This is not appropriate for small samples. We assume the underlying population variances are
equal: s12= s22 we estimate population variance as a weighted average of s 12 and s22 where the
weights are numbers of degrees of freedom in each sample.
One we have our estimate for population variance we can then use it to determine standard error
of the difference between two sample means, i.e we get an equation for the estimate standard
error of
The null hypotheses in this case is
An example will help make this clearer:

A company investigates two programmes for improving the sensitivity of its managers. One was a
more informal one whereas the second involved more formal classroom instruction. The informal
programme is more expensive and the president wants to know at the .05 level of significance
whether this expenditure has resulted in greater sensitivity.
8/19/2012 12:08 AM
SYLLABUS
25 of 71
12Managers were observed for the first method and 15 for the second. The sample data is as
follows:
The next step is to calculate estimate of the population variance :
We then calculate the t statistic for the difference between two means:
since it is a one tailed test at the .05 level of significance we look in the .1 column against 25
degrees of freedom.
t. critical at .05 level of significance= 1.708
Since calculated t< t critical , we accept the null hypothesis that the first method is significantly
superior to the second.
1.16 Summary
We now turn to case where we wish to compare the parameters for two different populations and
determine whether these differ from each other. In this case we are not really interested in the
actual value of the two parameters but the relation between the two parameter, i.e. is there a
significant difference between them.
1.17 Questions
1. What is Degree of Freedom?
8/19/2012 12:08 AM
SYLLABUS
26 of 71
2. Give Characteristics of the t distribution.
1.18 References
Boyd, westfall, and stasch, Marketing Research Text and Cases, All India Traveller Bookseller, New Delhi.
Brown, F.E. Marketing Research, a structure for decision making, Addison Wesley publishing company
Kothari, C.R. Research Methodology-Methods and Techniques, Wiley Eastern Ltd.
Stockton and Clark, Introduction to Business and Economic Statistics, D.B.Taraporevala Sons andCo.
Private Limited, Bombay.
Dunn Olive Jean and Virginia A Clarck, Applied Statistics John Wiley and Sons.
Green Paul E and Donald S. Tull, Research for Marketing Decisions Prentice Hall of India, New Delhi
8/19/2012 12:08 AM
SYLLABUS
27 of 71
Unit 2 Non Parametric Tests

Structure
2.0 Objectives
2.1 Introduction
2.2 Non Parametric Tests
2.3 Important Types of Nonparametric Tests
2.4 Non-parametric vs. Distribution-free Tests
2.5 Standard uses of Non Parametric Tests
2.6 Tests for Ordinal Data
2.7 The Null Hypothesis
2.8 Summary
2.9 Keywords
2.10 Questions
2.11 References
8/19/2012 12:08 AM
SYLLABUS
28 of 71
2.0 Objective
After studying this unit you will be to:
Define Non Parametric Tests
Discuss Important Types of Nonparametric Tests
Explain Non-parametric vs. Distribution-free Tests
2.1 Introduction
Let us understand that So far we have discussed a variety of tests that make inferences about a
population parameter such as the mean or the population proportion. These are termed
parametric tests and use parametric statistics from samples that come from the population being
tested. To use these tests we make several restrictive assumptions about the populations from
which we drew our samples. For example we assumed that our underlying population is normal.
However underlying populations are not always normal and in these situations we need to use
tests, which are not parametric.
Many such tests have been developed which do not make restrictive assumptions about the
shape of the population distribution. These are known as non-parametric or distribution free tests.
There are many such tests, we shall learn about some of the more popular ones.
2.2 Non Parametric Tests

What are Non Parametric Tests?
Statistical tests that do not require the estimate of population variance or mean and do not state
hypotheses about parameters are considered non-parametric tests. When do you use
non-parametric tests? Non-parametric tests are appropriately used when one or more of the
assumptions underlying a particular parametric test has been violated (generally normality or
homoscedasticity). Generally, however the t-test is fairly robust to all but the severest deviations
from the assumptions.
How do you know if the data are normally distributed? There are several techniques are generally
used to find out if a population has an underlying normal distribution. these include: goodness of
fit (low power), graphical assessment (not quantitative), Shapiro-Wilks W (n<50), or DAgostinoPearson Test K2 (preferred).
As noted non parametric tests do not make the assumption of normality about the population
distribution. The hypotheses of a non-parametric test are concerned with something other than
the value of a population parameter.
The main advantage of non-parametric methods is that they do not require that the underlying
population have a normal or any other shaped distribution. The main disadvantages of these tests
are that they ignore a certain amount of information. For example to convert data to non
parametric data we can convert numerical data to non parametric form by replacing numerical
8/19/2012 12:08 AM
SYLLABUS
29 of 71
values such as 113.45, 189.42, 76.5, 101.79 by either ascending or descending order ranks.
Therefore we can replace them by 1,2, 3, 4, and 5. How ever if we represent 189.42 by 5, we lose
some information, which is contained in the value 189.42. 189.42 are the largest value and this
represented by the rank 5. However the rank 5 could also represent 1189.42, as that would also
be the largest value. Therefore use of ranked data leads to some loss of information.
The second disadvantage is that these tests are not as sharp or efficient as parametric tests. The
estimate of an interval using a non-parametric test may be twice as large as for the parametric
case.
When we use nonparametric tests we trade off sharpness in estimation with the ability to make do
with less information and to calculate faster.
What happens when we use the wrong test in the wrong situation?
Generally, parametric tests are more powerful than non-parametric tests (e.g., the non-parametric
method will have a greater probability of committing a Type II error - accepting a false null
hypothesis)
2.3 Important Types of Nonparametric Tests:

Since the theory behind these tests is beyond the scope of our course we shall look at relevant
applications and methodologies for carrying out some of the more important non parametric tests.
Non-parametric tests are frequently used to test hypotheses with dependent samples.
The dependent sample tests are:
The signed test of paired data where positive and negative signs are substituted for
quantitative values.
Mcnemar test
Cochran
Wilcoxon test
Non-Parametric Tests for Independent Samples are:
Chi square test
Kolmogorov Smirnov one sample test
Of these tests we shall cover the McNemar test, the Mann Whitney U test, the Kologomorov
smironov test and the Wilcoxon test.
Table of equivalent parametric and non-parametric tests:
8/19/2012 12:08 AM
SYLLABUS
30 of 71
When do We use Non Parametric Tests?

We use non-parametric tests in least one of the following five types of situations :
The data entering the analysis are enumerative; that is, counted data represent the number of
observations in each category or cross-category.
The data are measured and/or analyzed using a nominal scale of measurement.
The data are measured and/or analyzed using an ordinal scale of measurement.
The inference does not concern a parameter in the population distribution; for example, the
hypothesis that a time-ordered set of observations exhibits a random pattern.
The probability distribution of the statistic upon which the analysis is based is not dependent
upon specific information or conditions (i.e., assumptions) about the population(s) from which
the sample(s) are drawn, but only upon general assumptions, such as a continuous and/or
symmetric population distribution.
According to these criteria, the distinction of non-parametric is accorded either because of the
The level of measurement used or required for the analysis, as in types 1,2, 3 . That is we use
either counted or ordinal or nominal scale data.
The type of inference, as in type 4. We do not make inferences about population parameters
such as the mean.
The generality of the assumptions made about the population distribution, as in type 5. That is
we do not know or make assumptions about the specific form of the underlying population
distribution.
2.4 Non-parametric vs. Distribution-free Tests:

As we have seen non-parametric tests are those used when some specific conditions for ordinary
tests are violated? Distribution-free tests are those for which the procedure is valid for all different
shapes of the population distribution.
For example, the Chi-square test concerning the variance of a given population is parametric
since this test requires that the population distribution be normal. On the other hand the
Chi-square test of independence does not assume normality condition, or even that the data are
numerical. The Kolmogorov-Smirnov test is a distribution-free test, which is applicable to
comparing two populations with any distribution of a continuous random variable.
2.5 Standard uses of Non Parametric Tests

Mann-Whitney
To be used with two independent groups (analogous to the independent groups t-test) we may
8/19/2012 12:08 AM
SYLLABUS
31 of 71
use the Mann-Whitney Rank Test as a non-parametric alternative to Students T-test when one
does not have normally distributed data.
Wilcoxon
To be used with two related (i.e., matched or repeated) groups (analogous to the dependent
samples t-test)
Kruskall-Wallis
To be used with two or more independent groups (analogous to the single-factor betweensubjects ANOVA)
Friedman
To be used with two or more related groups (analogous to the single-factor within-subjects
ANOVA)
We now look at a few of the non parametric tests in more details including their applications.
McNemar Test
This test is used for analyzing research designs of the before and after format where the data are
measured nominally. The samples therefore become dependent or related samples. The use of
this test is limited to the case where a 2x2 contingency table is involved. The test is most
popularly used to test response to a pre and post situation of a control group.
We can illustrate its use with an example:
A survey of 260 consumers was taken to test the effectiveness of mailed coupons and its effect on
individuals who changed their purchase rate for the product. The researcher took a random
sample of consumers before the release of the coupons to assess their purchase rate. On the
basis of their responses they were divided into groups as to their purchase rate (low, high). After
the campaign they were again asked to complete the forms and again classified on their
purchase rate. Table 1 shows the results from our sample.
8/19/2012 12:08 AM
SYLLABUS
32 of 71
Cases that showed a change in the before and after the campaign in terms of their purchase
response were placed in cells A and D. this was done as follows:
An individual is placed in cell A if he or she changed from a low purchase to a high purchase
rate.
Similarly hes placed in D if he goes from high to a low rate.
If no change is observed in his rate he is placed on cells BorC.
The researches wishes to determine if the mail order campaign was a success. We shall now
briefly outline the various steps involved in applying the McNemar test.
Step1
We state the null hypotheses. This essentially states that there is no perceptible or significant
change in purchase behavior of individuals. Thus for individuals who change their purchase rate
this means that the probability of those changing from high to low equals low to high. This equal
to .5.
Ho: P(A)=P(D)
Ha: P (A)P(D)
To test the null hypotheses we would examine the cases of change from cells A to D.
Step2
The level of significance is chosen, for example = .05
Step3
We now have to decide on the appropriate test to be used. The McNemar test is appropriate
because the study is a before and after study and the data are nominal. The study involves the
study of two related variables.
The McNemar test involves calculating the chi square value as given by the formula below:
That is we calculate the absolute difference between the A and D cells.

Step 4
The decision rule
For a=. 05, the critical value of 2 is 3.84 for degree of freedom 1. Therefore we will reject the null
hypotheses if the calculated 2 exceeds the critical value from the tables.
Step 5
8/19/2012 12:08 AM
SYLLABUS
33 of 71
We now actually calculate the test statistic:

The calculated ?2 =
Step 6
Draw a statistical conclusion.
Since calculated ?2 exceeds the critical value, we reject the null hypotheses and we can infer that
the mail coupon campaign was successful in increasing the purchase rate of the product under
study.
When an analysis involves more than two variables we use the Cochran Q test. For situations
where involving repeated observations where the dependent variable can take on only two values;
either 0 or 1.
2.6 Tests for Ordinal Data

So far the test we have discussed is applicable only to nominal data. We now look at a test,
which is specifically designed for ordinal data.
Kolmogorov Smirnov one Sample Test
This test is similar to the chi square test of goodness of fit. This is because it looks at the degree
of agreement between the distribution of observed values and some specified theoretical
distribution (expected frequencies). The Kolmogorov Smirov test is used if we want to compare
the distribution on an ordinal scale.
We can look at an example to see how this test works:
A paint manufacturer is interested in testing for four different shades of a colour: very light, very
light, bright and dark. Each respondent is shown four prints of the shades and asked to indicate
his preference. If colour shade is unimportant, the photos of each shade should be chosen
equally. Except for random differences. If colour shades are important the respondents should
prefer one of the extreme shades.
Since shade represents a natural ordering, the Kolmogorov test is applied to test the preference
hypothesis. The test involves specifying the cumulative frequency distribution. We then refer to
the sampling distribution indicates where there is a divergence between the two distributions is
likely to occur due to chance or if the observed difference is due to a result of preference.
Suppose for example that a sample of 200 homeowners and we got the following shade
preference distribution;
Very light: 80
Bright: 40
Dark: 20
8/19/2012 12:08 AM
SYLLABUS
34 of 71
The manufacturer asks whether these results indicate a preference.

The data is shown in table 2 below.
We would carry out the test as follows:

Specification of null hypotheses; Ho is that there would be no difference among the shades.
Ha: there is a difference in the shades of the new colour.
2. The level of significance: The test would be conducted at the 55 level.
Decision regarding statistical test. Here the Kolmogorov t-Smirnov test is appropriate because
the data measured are ordinal and we are interested in comparing the above frequency
distribution with a theoretical distribution.
This test focuses on the largest value of the deviations among observed and theoretical
proportions.
D= max |Fo (X)-Sn (X)|
Where Fo (X) is the specified cumulative frequency distribution under Ho for any value of x and is
the proportion of cases expected to have scores equal to or less than x.
So(X): observed cumulative frequency distribution of a random sample of N observations where X
is any possible score.
The decision rule
If the researcher chooses a =. 05, the critical value of D for large samples is given by the
formula 1.36n where n is the sample size. For example the critical value is .96. the decion rule
8/19/2012 12:08 AM
SYLLABUS
35 of 71
therefore states that null hypotheses will be rejected if computed D >.96.

Calculation of test Statistic
The theoretical probability distribution is calculated by taking each class frequency as if it were
under the null hypotheses, i.e., in our case they would be equal. It is then expressed as a fraction
of the total sample. Thus row 2 would be calculated as 50/ 200=. 25, 100/200=. 5, 150/200=. 75,
200/200=1
The observed cumulative frequency of row 3 is found by 80/ 200=. 40, 140/200=. 70, 180/200=. 90
The calculated D value is the point of greatest divergence between cumulative observed and
theoretical distributions. In our example this is .20
Drawing a statistical conclusion: Since calculated D value (.20)
exceeds the critical value of .096, the null hypotheses of no
difference among shades are quickly rejected.
Mann Whitney U Test
This is used for data, which is ordinal and can be ranked. This test makes use of the actual ranks
of the observations as a means of testing the hypotheses about the identity of two population
distributions.
An Example will Illustrate this Test
To illustrate its use we will examine regular and commercial account satisfaction data from Table
3. The table contains the attitude scores obtained from 30 customers (15 regular accounts and 15
commercial accounts). The scores from the combined samples were then ranked in terms of their
magnitude of the original score (columns 3 and 4 of Table3). That is, the highest score gets rank
1, the next highest 2, and so on.
2.7 The Null Hypothesis

Ho is that there is no difference in attitudes of the two groups of accounts towards bank services..
Ha is that there is a significant difference difference in attitudes of the two groups.
The level of significance for this test is .05.
8/19/2012 12:08 AM
SYLLABUS
36 of 71
The Mann whiteny Test is used because the data is ordinal and converted into ranks.. Also the
samples are independent.
The formula for the Mann whitney U value is :
Analysis of Differences
Where n1 and n2 are the two sample sizes and R1 and R2are the sums of the ranks for each
group. Letting regular accounts be sample 1 and the commercial accounts be sample 2, we
ind that
n1 = 15 n2=15 R1 = 198 R2= 267
The critical value for the statistic U* is found in Appendix I. For on = 0.05, n1 = 15 and n2 = 15,
the critical value for the Mann-Whitney statistic is U* = 64 for a two-tailed test. For this test, the
null hypothesis will be rejected if the computed value, U, is 64 or less. Otherwise, it will not be
rejected. This decision is just the opposite of the decision making procedure we followed for most
of the other tests of significance.
Calculation of the Test Statistic.
Therefore,
U1 = (15)(15) + 15(15 + 1) - 198
2
8/19/2012 12:08 AM
SYLLABUS
37 of 71
= 225 + 120 - 198

= 345 - 198
= 147 And U2 = (15)(15) - 147 = .225 - 147
= 78
The Mann-Whitney U is the smaller of the two U values calculated. In this case the Mann-Whitney
U is 78. Inspection of the formulas for calculating U will indicate that the more similar the two
groups are in their attitudes, the smaller the R values will be and the larger the value of U.
Therefore we are testing the probability of obtaining a value the size of the smallest of the two
groups if the two groups are indeed similar in their attitudes.
Drawing a Statistical Conclusion:
Since the computed U is larger than the critical U*, it does not fall in the critical region and the
null hypothesis is not rejected. For this test the computed value must be less than the critical
value U* to reject the null hypothesis. Once again, the evidence does not support a difference in
the attitudes between the two groups.
The Kruskal- Wallis test is an extension of the Mann-Whitney U t situations where more than two
independent samples are being compared example, this test could be used if the customers were
divided into three or groups based on some criterion, such as regular accounts, commercial aCC4
and charge accounts.
Signed and Rank or Wilcoxon test
This test is the complement of the Mann Whitney U test and is used when ordinal data on two
samples are involved and the two samples are related or dependent. This test is therefore suited
to a pre-test and post-test situation. The test can also be used for interval or ratio data when the
assumptions underlying the parametric z or t test cannot be met.
An example illustrates this rule:
To illustrate the procedure, suppose that our bank in the previous examples wanted to test the
effectiveness of an ad campaign intended to enhance the awareness of the banks service
features. The bank administered a questionnaire before the ad campaign designed to measure
the awareness of services offered. After the ad campaign, the bank administered the same
questionnaire to the same group of people. The bank wished to determine whether there was any
change in the awareness of services offered due to the ad campaign. Both the before-and after
ad campaign scores is presented in Table 4
8/19/2012 12:08 AM
SYLLABUS
38 of 71
Solution
Step 1
The Null Hypothesis. The null hypothesis to be tested is that there is no difference in awareness
of services offered after the ad campaign. The alternative hyp6thesis would be that there was an
awareness of the services after the ad campaign.
Step 2The Level of Significance
It was decided that = 0.05. Step J.
The Statistical Test
The Wilcoxon test is appropriate because the study is of related samples and in which the data
measured, is ordinal and the differences can be ranked in magnitude. The test statistic calculated
is the T value. Since the direction of the difference is predicted, a one-tailed test is appropriate.
Step 3. The Decision Rule
The critical value of the Wilcoxon T is found Appendix J for n = 10 at the 0.05 level of significance
and a one-toiled test are 10. This indicates that a computed Tvalue of less than 10, the critical
value, rejects the null hypothesis. The argument is similar to that, of the Mann-Whitney U
statistic.
Step 4. Calculate the Test Statistic
The procedure for the test is very simple. The signed difference between each pair of
observations is found. Then these differences are ranked-ordered without regard to their
algebraic sign. Finally, the sign of the difference is attached to the rank for that difference. The
test statistic, T, is the smaller of the two sums of the ranks. For our example, T = 6.5 since the
smaller sum is associated with the negative difference.
If the null hypothesis is true the sums of positive and negative ranks should be approximately
equal. However the larger the difference between the underlying population, the smaller would
be the value of T since it is defined as the smaller of the ranks.
Step5 We draw a statistical conclusion; since the computed T value is 6.5 is less than the critical
T valie of 120; the null hypothesis, which states that there is no difference in the awareness of
bank services, is rejected.
8/19/2012 12:08 AM
SYLLABUS
39 of 71
2.8 Summary
The main advantage of non-parametric methods is that they do not require that the underlying
population have a normal or any other shaped distribution. The researches wishes to determine if
the mail order campaign was a success. We shall now briefly outline the various steps involved in
applying the McNemar test.
2.9 Keywords
McNemar Test
This test is used for analyzing research designs of the before and after format where the data are
measured nominally.
Kolmogorov Smirnov one Sample Test
This test is similar to the chi square test of goodness of fit.
Signed and Rank or Wilcoxon test
This test is the complement of the Mann Whitney U test and is used when ordinal data on two
samples are involved and the two samples are related or dependent.
2.10 Questions
1. What are Non Parametric Tests?
2. What is the Theoretical aspect of the t distribution?
2.11 References
8/19/2012 12:08 AM
SYLLABUS
40 of 71
Unit 3 Chi-Square Test

Structure
3.0 Objectives
3.1 Introduction
3.2 Key Definitions
3.3 Properties of the Chi-Square
3.4 Chi-Square Probabilities
3.5 Degree of Freedom Which Arent in the Table
3.6 Uses of Chi-square Test
3.7 Applications of Chi-square Test
3.8 Summary
3.9 Keywords
3.10 Questions
3.11 References
8/19/2012 12:08 AM
SYLLABUS
41 of 71
3.0 Objective
After studying this unit you will be able to:
Define Properties of the Chi-Square
Discuss Chi-Square Probabilities
Describe Uses of Chi-square Test
3.1 Introduction
Let us understand that chi-square test is for goodness of fit and independence of attributes
The Chi Square distribution is a mathematical distribution that is used directly or indirectly in
many tests of significance. The most common use of the chi square distribution is to test
differences between proportions. Although this test is by no means the only test based on the chi
square distribution, it has come to be known as the chi square test. The chi square distribution
has one parameter, its degrees of freedom (df). It has a positive skew; the skew is less with more
degrees of freedom. The mean of a chi square distribution is its df. The mode is df - 2 and the
median is approximately df -0 .7.
3.2 Key Definitions

Chi-square Distribution
A distribution obtained from the multiplying the ratio of sample variance to population variance by
the degrees of freedom when random samples are selected from a normally distributed
population
Contingency Table
Data arranged in table form for the chi-square independence test
Expected Frequency
The frequencies obtained by calculation.
Goodness-of-fit Test
A test to see if a sample comes from a population with the given distribution.
Independence Test
A test to see if the row and column variables are independent.
Observed Frequency
8/19/2012 12:08 AM
SYLLABUS
42 of 71
The frequencies obtained by observation. These are the sample frequencies.

Chi-Square Distribution
The chi-square ? distribution is obtained from the values of the ratio of the
Sample variance and population variance multiplied by the degrees of freedom.
This occurs when the population is normally distributed with population variance s 2.
Symbolically, Chi-square is defined as ,(with usual notations)
3.3 Properties of the Chi-Square

Chi-square is non-negative. Since it is the ratio of two non-negative values, therefore must be
non-negative itself.
Chi-square is non-symmetric.
For each degree of freedom, we have one chi-square distributions
The degree of freedom when working with a single population variance is n-1.
3.4 Chi-Square Probabilities

Since the chi-square distribution isnt symmetric, the method for looking up left-tail values is
different from the method for looking up right tail values.
Area to the right - just use the area given.
Area to the left - the table requires the area to the right, so subtract the given area from one
and look this area up in the table.
Area in both tails - divide the area by two. Look up this area for the right critical value and one
minus this area for the left critical value.
3.5 Degree of Freedom Which Arent in the Table

When the degrees of freedom arent listed in the table, there are a couple of choices that you
have.
You can interpolate.
1. This is probably the more accurate way.
2. Interpolation involves estimating the critical value by figuring how far the given degrees of
freedom are between the two df in the table and going that far between the critical values
in the table.
For Example
Most people born in the 70s didnt have to learn interpolation in high school because they had
calculators which would do logarithms (we had to use tables in the good old days).
8/19/2012 12:08 AM
SYLLABUS
43 of 71
You can go with the critical value which is less likely to cause you to reject in error (type I
error).
For a right tail test, this is the critical value further to the right (larger).
For a left tail test, it is the value further to the left (smaller).
For a two-tail test, its the value further to the left and the value further to the right. Note, it is
not the column with the degrees of freedom further to the right, its the critical value which is
further to the right.
3.6 Uses of Chi-square Test

1.Goodness-of-fit Test
The idea behind the chi-square goodness-of-fit test is to see if the sample comes from the
population with the claimed distribution.
Another way of looking at that is to ask if the frequency distribution fits a specific pattern.
Two values are involved, an observed value, which is the frequency of a category from a sample,
and the expected frequency, which is calculated based upon the claimed distribution.
The idea is that if the observed frequency is really close to the claimed (expected) frequency,
then the square of the deviations will be small.
The square of the deviation is divided by the expected frequency to weight frequencies.
A difference of 10 may be very significant if 12 was the expected frequency, but a difference of 10
isnt very significant at all if the expected frequency was 1200.
If the sum of these weighted squared deviations is small, the observed frequencies are close to
the expected frequencies and there would be no reason to reject the claim that it came from that
distribution. Only when the sum is large is the reason to question the distribution. Therefore, the
chi-square goodness-of-fit test is always a right tail test.
The test statistic has a chi-square distribution when the following assumptions are met
The data are obtained from a random sample
The expected frequency of each category must be at least 5. This goes back to the
requirement that the data be normally distributed. Youre simulating a multinomial experiment
(using a discrete distribution) with the goodness-of-fit test (and a continuous distribution), and
if each expected frequency is at least five then you can use the normal distribution to
approximate (much like the binomial).
The following are properties of the goodness-of-fit test
8/19/2012 12:08 AM
SYLLABUS
44 of 71
The data are the observed frequencies. This means that there is only one data value for each
category. Therefore,
The degrees of freedom is one less than the number of categories, not one less than the
sample size.
It is always a right tail test.
It has a chi-square distribution.
The value of the test statistic doesnt change if the order
of the categories is switched.
2. Test for Independence
In the test for independence, the claim is that the row and column variables are independent of
each other.
This is the null hypothesis.
The multiplication rule said that if two events were independent, then the probability of both
occurring was the product of the probabilities of each occurring. This is key to working the test for
independence.
If you end up rejecting the null hypothesis, then the assumption must have been wrong and the
row and column variable are dependent. Remember, all hypothesis testing is done under the
assumption the null hypothesis is true.
The test statistic used is the same as the chi-square goodness-of-fit test. The principle behind the
test for independence is the same as the principle behind the goodness-of-fit test.
The Test for Independence is Always a Right Tail Test
In fact, you can think of the test for independence as a goodness-of-fit test where the data is
arranged into table form. This table is called a contingency table.
The test statistic has a chi-square distribution when the following assumptions are met
The data are obtained from a random sample
The expected frequency of each category must be at least 5.
The following are properties of the test for independence

The data are the observed frequencies.
8/19/2012 12:08 AM
SYLLABUS
45 of 71
The data is arranged into a contingency table.

The degrees of freedom are the degrees of freedom for the row variable times the degrees of
freedom for the column variable. It is not one less than the sample size, it is the product of the
two degrees of freedom.
It is always a right tail test.
It has a chi-square distribution.
The expected value is computed by taking the row total times the column total and dividing by
the grand total
The value of the test statistic doesnt change if the order of the rows or columns are switched.
The value of the test statistic doesnt change if the rows and columns are interchanged
(transpose of the matrix)
3.7 Applications of Chi-square Test

I. Goodness of fit:
Habitat Use. If you are looking for habitat selection or avoidance in a species (e.g., black bear),
you can use a goodness of fit test to see if the animals are using habitat in proportion to its
availability. If you have collected 100 radio-locations on a bear or bears, you would expect that, if
the species were not selecting or avoiding habitat, you would find the radio-locations spread in
each habitat type depending on its general availability (If 90% of the area was lowland conifer,
you would expect 90% of the locations to occur in that habitat type). Imagine you generate the
following data from your spring bear study:
Ho : The bears are using habitat in proportion to its availability.
Fill in the expected number of radio-locations in each cover type. 2. 3.

Calculate the chi-square value for these bears.
Given that the degrees of freedom are (r-1)*(c-1)= 4, the critical value for chi-square (at alpha
= 0.05) is anything greater than 9.49. Would you accept or reject the null hypothesis? What
does that mean in laymans terms?
Survival. Suppose you are investigating the population dynamics of the deer herd at Sandhill and
hypothesize that a constant death rate of 50% per year has existed over the past several years
with a stable herd. You sample the population at random and classify deer into age groups as
indicated:
8/19/2012 12:08 AM
SYLLABUS
46 of 71
The expected value, (X), of the first age group is obtained from the formula:
Subsequent expected values are computed by applying the expected 50% death rate (d) for each
succeeding year.
State the null hypothesis
Calculate the Chi-square value.
Knowing the critical value for 5 degrees of freedom (* = 0.05) is 11.0705, what do you
conclude about the fit between the observed and hypothesized death rates? Are there
significant differences??
Model Selection
Many programs develop predictive equations for data sets. A common test for the fit of a model to
the data is chi-square goodness-of-fit test. For example, program DISTANCE, which develops
curves to estimate probabilities of detection, uses discrete distance categories (e.g., 0-5m, 5-10m,
etc.) to see how well the model predicts the number of objects that should be seen in each
distance category (Expected) versus what the data actually show (Observed). II)
Contingency Tables and Tests for Independence of Factors:
Survival
Is the probability of being male or female independent of being alive or dead? Lets use data on
ruffed grouse. You collect mortality data from 100 birds you radio-collared and test the following
hypothesis:
Ho: The 2 sets of attributes (death and sex of bird) are unrelated (independent).
8/19/2012 12:08 AM
SYLLABUS
47 of 71
Expected values for each cell can be can be calculated by multiplying the row total by the column
total and dividing by the grand total.
EX: Expected Value = 70 * 67 / 100 = 46.9
Calculate the chi-square value.
Knowing the critical value for 1 degree of freedom (alpha = 0.05) is anything greater than
3.84146, what can you conclude about the independence of these 2 factors? rates?
II Chi Square Test for Independence (2-Way chi-square)
A large-scale national randomized experiment was conducted in the 1980s to see if daily aspirin
consumption (as compared to an identical, but inert placebo) would reduce the rate of heart
attacks. This study (The Physicians Health Study) was described in one of the episodes of the
statistics video series A Against All Odds.@ Here are the actual results from the study using
22,071 doctors who were followed for 5 years:
A chi-square analysis indicated that there was a significant relationship between aspirin condition
and incidence of heart attacks, chi-square (1, N=22,071)= 25.01, p<.001. A greater percentage of
heart attacks occurred for participants taking the placebo (M=1.7%) compared to those taking
aspirin (M=0.
Application of chi-square test
Product and Process Comparisons
8/19/2012 12:08 AM
SYLLABUS
48 of 71
Comparisons based on data from more than two processes

How can we compare the results of classifying according to several categories
Contingency Table approach
When items are classified according to two or more criteria, it is often of interest to decide whether
these criteria act independently of one another.
For example, suppose we wish to classify defects found in wafers produced in a manufacturing
plant, first according to the type of defect and, second, according to the production shift during
which the wafers were produced. If the proportions of the various types of defects are constant
from shift to shift, then classification by defects is independent of the classification by production
shift. On the other hand, if the proportions of the various defects vary from shift to shift, then the
classification by defects depends upon or is contingent upon the shift classification and the
classifications are dependent.
In the process of investigating whether one method of classification is contingent upon another, it
is customary to display the data by using a cross classification in an array consisting of r rows
and c columns called a contingency table. A contingency table consists of r x c cells representing
the r x c possible outcomes in the classification process.
Let us construct an industrial case Industrial example A total of 309 wafer defects were recorded
and the defects were classified as being one of four types, A, B, C, or D. At the same time each
wafer was identified according to the production shift in which it was manufactured, 1, 2, or 3.
Contingency table classifying defects in wafers according to type and production shift
These counts are presented in the following table
(Note: the numbers in parentheses are the expected cell frequencies). Column probabilities Let p
A
be the probability that a defect will be of type A. Likewise, define p B, pC, and pD as the
8/19/2012 12:08 AM
SYLLABUS
49 of 71
probabilities of observing the other three types of defects. These probabilities, which are called
the column probabilities, will satisfy the requirement
Row probabilities By the same token, let pi (i=1, 2, or 3) be the row probability that a defect will
have occurred during shift i, where
Multiplicative Law of Probability

Then if the two classifications are independent of each other, a cell probability will equal the
product of its respective row and column probabilities in accordance with the Multiplicative Law of
Probability
Example of obtaining column and row probabilities
For example, the probability that a particular defect will occur in shift 1 and is of type A is (p 1)
(p A). While the numerical values of the cell probabilities are unspecified, the null hypothesis
states that each cell probability will equal the product of its respective row and column
probabilities. This condition implies independence of the two classifications. The alternative
hypothesis is that this equality does not hold for at least one cell.
In other words, we state the null hypothesis as H0: the two classifications are independent, while
the alternative hypothesis is Ha: the classifications are dependent.
To obtain the observed column probability, divide the column total by the grand total, n. Denoting
the total of column j as cj, we get.
Similarly, the row probabilities p 1, p2, and p 3 are estimated by dividing the row totals r1, r 2, and
r3 by the
Total n, respectively
Denote the observed frequency of the cell in row i and column jof the contingency table by n...
Then we have
8/19/2012 12:08 AM
SYLLABUS
50 of 71
Expected cell frequencies

Denote the observed frequency of the cell in row i and column jof the contingency table by n...
Then we have
Estimated expected cell frequency when H0 is true.

In other words, when the row and column classifications are independent, the estimated expected
value of the observed cell frequency ni in an r x c contingency table is equal to its respective row
and column totals divided by the total frequency.
The estimated cell frequencies are shown in parentheses in the contingency table above
1. Chi-Square (c2) Analysis- Introduction
Consider the following decision situations:
1. Are all package designs equally preferred? 2) Are all brands equally preferred? 3) Is their any
association between income level and brand preference? 4) Is their any association between
family size and size of washing machine bought? 5) Are the attributes educational background
and type of job chosen independent? The answer to these questions require the help of
Chi-Square (c2) analysis. The first two questions can be unfolded using Chi-Square test of
goodness of fit for a single variable while solution to questions 3, 4, and 5 need the help of
Chi-Square test of independence in a contingency table. Please note that the variables involved
in Chi-Square analysis are nominally scaled. Nominal data are also known by two namescategorical data and attribute data.The symbol c2 used here is to denote the chi-square
distribution whose value depends upon the number of degrees of freedom (d.f). As we know,
chi-square distribution is a skewed distribution particularly with smaller d.f. As the sample size
and therefore the d.f. increases and becomes large, the c2 distribution approaches normality.
c2 tests are nonparametric or distribution-free in nature. This means that no assumption needs to
be made about the form of the original population distribution from which the samples are drawn.
Please note that all parametric tests make the assumption that the samples are drawn from a
8/19/2012 12:08 AM
SYLLABUS
51 of 71
specified or assumed distribution such as the normal distribution.

For a meaningful appreciation of the conditions/assumptions involved in using chi-square
analysis, please go through the contents of hyperstat on chi-square test meticulously.
2. Chi-Square Test-Goodness of Fit
A number of marketing problems involve decision situations in which it is important for a
marketing manager to know whether the pattern of frequencies that are observed fit well with the
expected ones. The appropriate test is the c2 test of goodness of fit. The illustration given below
will clarify the role of c2 in which only one categorical variable is involved.
Problem
In consumer marketing, a common problem that any marketing manager faces is the selection of
appropriate colors for package design. Assume that a marketing manager wishes to compare five
different colors of package design. He is interested in knowing which of the five is the most
preferred one so that it can be introduced in the market. A random sample of 400 consumers
reveals the following:
Do the consumer preferences for package colors show any significant difference?
Solution
If you look at the data, you may be tempted to infer that Blue is the most preferred color.
Statistically, you have to find out whether this preference could have arisen due to chance. The
appropriate test statistic is the c2 test of goodness of fit.
Null Hypothesis: All colors are equally preferred.
Alternative Hypothesis: They are not equally preferred.
8/19/2012 12:08 AM
SYLLABUS
52 of 71
Please note that under the null hypothesis of equal preference for all colors being true, the
expected frequencies for all the colors will be equal to 80. Applying the formula
We get the computed value of chi-square ( c2 ) = 11.400

The critical value of c2 at 5% level of significance for 4 degrees of freedom is 9.488. So, the null
hypothesis is rejected. The inference is that all colors are not equally preferred by the consumers.
In particular, Blue is the most preferred one. The marketing manager can introduce blue color
package in the market.
3. Chi-Square Test of Independence
The goodness-of-fit test discussed above is appropriate for situations that involve one categorical
variable. If there are two categorical variables, and our interest is to examine whether these two
variables are associated with each other, the chi-square( c2 ) test of independence is the correct
tool to use. This test is very popular in analyzing cross-tabulations in which an investigator is
keen to find out whether the two attributes of interest have any relationship with each other.
The cross-tabulation is popularly called by the term contingency table. It contains frequency data
that correspond to the categorical variables in the row and column. The marginal totals of the
rows and columns are used to calculate the expected frequencies that will be part of the
computation of the c2 statistic. For calculations on expected frequencies, refer hyperstat on c2
test.
3.8 Summary
A number of marketing problems involve decision situations in which it is important for a
marketing manager to know whether the pattern of frequencies that are observed fit well with the
expected ones.In consumer marketing, a common problem that any marketing manager faces is
the selection of appropriate colors for package design.
8/19/2012 12:08 AM
SYLLABUS
53 of 71
3.9 Keywords
Chi-square Distribution: A distribution obtained from the multiplying the ratio of sample variance
to population variance by the degrees of freedom when random samples are selected from a
normally distributed population.
Contingency Table: Data arranged in table form for the chi-square independence test
Expected Frequency: The frequencies obtained by calculation.
Goodness-of-fit Test: A test to see if a sample comes from a population with the given
distribution.
Independence Test: A test to see if the row and column variables are independent.
Observed Frequency: The frequencies obtained by observation. These are the sample
frequencies.
3.10 Questions
1. What are the Properties of the Chi-Square?
2. Discuss Chi-Square Probabilities.
3. What is the Degree of Freedom Which Arent in the Table?
3.11 References
Unit 4 Analysis of Variance (ANOVA)

Structure
8/19/2012 12:08 AM
SYLLABUS
54 of 71
4.0 Objectives
4.1 Introduction
4.2 Logic behind ANova
4.3 Applications of Anova
4.4 Dependent and Independent Variables.
4.5 Summary
4.6 Keywords
4.7 Questions
4.8 References
8/19/2012 12:08 AM
SYLLABUS
55 of 71
4.0 Objectives
After studying this unit you will be to :
Define Logic behind ANova
Discuss Applications of Anova
Describe Dependent and Independent Variables.
4.1 Introduction
Let us understand that the tests we have learned up to this point allow us to test hypotheses that
examine the difference between only two means.
Analysis of Variance or Anova will allow us to test the difference between 2 or more means.
Anova does this by examining the ratio of variability between two conditions and variability
within each condition.
A t-test would compare the likelihood of observing the difference in the mean number of words
recalled for each group.
An ANOVA test, on the other hand, would compare the variability that we observe between the
two conditions to the variability observed within each condition. Recall that we measure variability
as the sum of the difference of each score from the mean. When we actually calculate an ANOVA
we will use a short-cut formula.
Thus, when the variability that we predict (between the two groups) is much greater than the
variability we dont predict (within each group), and then we will conclude that our treatments
produce different results.
X = the ith observation in the jth group
ij
n = the number of observations in group j
n = the total number of observations in all groups combined
c = the number of groups or levels
8/19/2012 12:08 AM
SYLLABUS
56 of 71
6. Accept null hypothesis or reject null hypothesis

7. There are no significant differences among the c means or there is at least one inequality
among the c means.
An Illustrative Numerical Example for Anova
Let us introduce the Anova in simplest forms by numerical illustration.
Example:
Consider the following (small, and integer, indeed for illustration while saving space) random
samples from three different populations.
Hypothesis:
Null Hypothesis, H0: 1 = 2 = 3,
Alternative Hypothesis,HA: at least two of the means are not equal.
(Not all mj are equal)
8/19/2012 12:08 AM
SYLLABUS
57 of 71
Significance level a= 0.05,

the critical value from F-table is : F 0.05, 2, 12 = 3.89.
Demonstrate that, SST= SSB+SSW

Computation of Sample SST:
With the grand mean = 3, first, start with taking the difference between each observation and the
grand mean, and then square it for each data point.
Therefore SST=36 with d.f = 15-1 = 14

Computation of Sample SSB:
Second, let all the data in each sample have the same value as the mean in that sample. This
removes any variation WITHIN. Compute SS differences from the grand mean.
Therefore SSB = 10, with d.f = 3-1 = 2

Computation of Sample SSW:
Third, compute the SS difference within each sample using their own sample means.
This provides SS deviation Within all samples.
8/19/2012 12:08 AM
SYLLABUS
58 of 71
SSW = 26 with d.f = 3(5-1) = 12

Results are: SST = SSB + SSW, and d.f = d.f + d.f , as expected.
SST
SB SSW
Now, construct the ANOVA table for this numerical example by plugging the results of your
computation in the Anova Table.
Conclusion: There is not enough evidence to reject the null hypothesis Ho.
4.2 Logic Behind ANova

First, let us try to explain the logic and then illustrate it with a simple example.
In performing Anova test, we are trying to determine if a certain number of population means are
equal.
To do that, we measure the difference of the sample means and compare that to the variability
within the sample observations. That is why the test statistic is the ratio of the between-sample
variation (MST) and the within-sample variation (MSE). If this ratio is close to 1, there is evidence
that the population means are equal.
Heres a hypothetical example:
Many people believe that men get paid more in the business world than women, simply because
they are male. To justify or reject such a claim, you could look at the variation within each group
(one group being womens salaries and the other being men salaries) and compare that to the
variation between the means of randomly selected samples of each population. If the variation in
the womens salaries is much larger than the variation between the men and womens mean
salaries, one could say that because the variation is so large within the womens group that this
may not be a gender-related problem.
8/19/2012 12:08 AM
SYLLABUS
59 of 71
Now, getting back to our numerical example, we notice that: given the test conclusion and the
Anova tests conditions, we may conclude that these three populations are in fact the same
population.
Therefore, the Anova technique could be used as a measuring tool and statistical routine for
quality control as described below using our numerical example.
Construction of the Control Chart for the Sample Means:
Under the null hypothesis the ANOVA concludes that 1 = 2 = 3; that is, we have a hypothetical
parent population.
The question is, what is its variance? The estimated variance is 36
/ 14 = 2.75.
Thus, estimated standard deviation is = 1.60 and estimated standard deviation for the means is
1.6 /^5 = 0.71.
Under the conditions of Anova, we can construct a control chart with the warning limits = 3
2(0.71); the action limits = 3 3(0.71). The following figure depicts the control chart.
8/19/2012 12:08 AM
SYLLABUS
60 of 71
Spss program for Anova: More Than Two Independent Means

$Spss/Output=4-1.out1 Title Analysis Of Variance - 1st Iteration Data List Free File=a.in/Gp Y
Oneway Y By Gp(1,5)/Ranges=Duncan /Statistics Descriptives Homogeneity Statistics 1
Manova Y By Gp(1,5) Print = Homogeneity (Bartlett) /
Npar Tests K-w Y By Gp(1,5)/
Finish
Anova like two population t-test can go wrong when the equality of variances condition is not met.
General Rule of 2, Homogeneity of Variance: Checking the equality of variances
For 3 or more populations, there is a practical rule known as the Rule of 2. According to this rule,
one divides the highest variance of a sample by the lowest variance of the other sample.
Given that the sample sizes are almost the same, and the value of this division is less than 2,
then, the variations of the populations are almost the same.
Homogeneity of Variance
Bartletts Test (normal distribution) Test Statistic: 0.033 P-Value : 0.984
8/19/2012 12:08 AM
SYLLABUS
61 of 71
One-way Analysis of Variance

Analysis of Variance for time
Source DF SS MS F P
method 2 98.40 49.20 6.38 0.026
Error 7 54.00 7.71
Total 9 152.40
Individual 95% CIs For Mean
Based on Pooled StDev
Tukeys pairwise comparisons

Family error rate = 0.0500
Individual error rate = 0.0214 Critical value = 4.17
Intervals for (column level mean) - (row level mean)
Example of How An Anova Should Be Written Up: Check of Assumption Of Equal Variances: H0:
The variances are equal. H1: The variances are not equal
Homogeneity of Variance
Bartletts Test (normal distribution)
Test Statistic : 0.033
P-Value : 0.984
8/19/2012 12:08 AM
SYLLABUS
62 of 71
Since the p-value = 0.984 > 0.05, we DNR Ho.

Therefore, there are no significant differences among the variances. The assumption of equal
variances is met.
Check The Assumption of Normality:
H0: The residuals fit a normal distribution.
H: The residuals do not fit a normal distribution
Since the p-value > 0.15 > 0.05, we DNR Ho. Therefore, the normality assumption is met. Check
for Differences Among The Means:
1.
H1: at least one mj not equal
Anova
a= 0.05
If the p-value < a reject H0.
If the p-value a do not reject H0.
One-way Analysis of Variance
Analysis of Variance for time
Individual 95% CIs For Mean

Based on Pooled StDev
8/19/2012 12:08 AM
SYLLABUS
63 of 71
Since the p-value = 0.026 < 0.05, we reject Ho.

Therefore, there is at least one significant difference in the average time it takes to complete the
tax return among the three methods.
Tukeys Pairwise Comparisons
Family error rate = 0.0500
Individual error rate = 0.0214 Critical value = 4.17
Intervals for (column level mean) - (row level mean)
12
-1.255 11.255
-9.255 -14.687
3.255 -1.313
Methods 2 and 3 are significantly different from each other. Method 1 is not significantly different
from either method 2 or 3. By examining the means displayed with the ANOVA analysis, I would
recommend Method 2 as the method tax prepares should use since the group using this method
requires significantly less time to prepare returns on average than the group using method 3.
4.3 Applications of Anova

The Purpose of Analysis of Variance
In general, the purpose of analysis of variance (Anova) is to test for significant differences
between means.
Elementary Concepts provides a brief introduction into the basics of statistical significance
testing. If we are only comparing two means, then Anova will give the same results as the t test
for independent samples (if we are comparing two different groups of cases or observations), or
the t test for dependent samples (if we are comparing two variables in one set of cases or
observations).
If you are not familiar with those tests you may at this point want to brush up on your knowledge
8/19/2012 12:08 AM
SYLLABUS
64 of 71
about those tests by reading Basic Statistics and Tables.

Why the name analysis of variance?
It may seem odd to you that a procedure that compares means is called analysis of variance.
However, this name is derived from the fact that in order to test for statistical significance between
means, we are actually comparing (i.e., analyzing) variances.
The Partitioning of Sums of Squares
Multi-Factor Anova
Interaction Effects
The Partioning of Sums of Squares
At the heart of Anova is the fact that variances can be divided up, that is, partitioned.
Remember that the variance is computed as the sum of squared deviations from the overall
mean, divided by n-1 (sample size minus one).
Thus, given a certain n, the variance is a function of the sums of (deviation) squares, or SS for
short. Partitioning of variance works as follows.
The means for the two groups are quite different (2 and 6, respectively).
The sums of squares within each group are equal to 2. Adding them together, we get 4.
If we now repeat these computations, ignoring group membership, that is, if we compute the total
SS based on the overall mean, we get the number 28. In other words, computing the variance
(sums of squares) based on the within-group variability yields a much smaller estimate of
variance than computing it based on the total variability (the overall mean). The reason for this in
the above example is of course that there is a large difference between means, and it is this
difference that accounts for the difference in the SS.
In fact, if we were to perform an ANOVA on the above data, we would get the following result:
8/19/2012 12:08 AM
SYLLABUS
65 of 71
SS Error and SS Effect.

The within-group variability (SS) is usually referred to as Error variance. This term denotes the
fact that we cannot readily explain or account for it in the current design. However, the SS Effect
we can explain.
Namely, it is due to the differences in means between the groups. Put another way, group
membership explains this variability because we know that it is due to the differences in means.
Significance Testing.
The basic idea of statistical significance testing is discussed in Elementary Concepts. Elementary
Concepts also explains why very many statistical test represent ratios of explained to unexplained
variability. ANova is a good example of this.
Here, we base this test on a comparison of the variance due to the between- groups variability
(called Mean Square Effect, or MSeffect) with the within- group variability (called Mean Square
Error, or Mserror; this term was first used by Edgeworth, 1885).
Under the null hypothesis (that there are no mean differences between groups in the population),
we would still expect some minor random fluctuation in the means for the two groups when taking
small samples (as in our example).
Therefore, under the null hypothesis, the variance estimated based on within-group variability
should be about the same as the variance due to between-groups variability. We can compare
those two estimates of variance via the F test (see also F Distribution), which tests whether the
ratio of the two variance estimates is significantly greater than 1.
In our example above, that test is highly significant, and we would in fact conclude that the
means for the two groups are significantly different from each other.
4.4 Dependent and Independent Variables.

The variables that are measured (e.g., a test score) are called dependent variables. The variables
that are manipulated or controlled (e.g., a teaching method or some other criterion used to divide
observations into groups that are compared) are called factors or independent variables
Multi-Factor ANova
In the simple example above, it may have occurred to you that we could have simply computed a
t test for independent samples to arrive at the same conclusion. And, indeed, we would get the
identical result if we were to compare the two groups using this test.
However, Anova is a much more flexible and powerful technique that can be applied to much
more complex research issues.
8/19/2012 12:08 AM
SYLLABUS
66 of 71
Multiple Factors
The world is complex and multivariate in nature, and instances when a single variable completely
explains a phenomenon are rare.
For example, when trying to explore how to grow a bigger tomato, we would need to consider
factors that have to do with the plants genetic makeup, soil conditions, lighting, temperature, etc.
Thus, in a typical experiment, many factors are taken into account. One important reason for
using ANOVA methods rather than multiple two-group studies analyzed via t tests is that the
former method is more efficient, and with fewer observations we can gain more information.
Let us expand on this statement.
Controlling for Factors
Suppose that in the above two-group example we introduce another grouping factor, for example,
Gender. Imagine that in each group we have 3 males and 3 females. We could summarize this
design in a 2 by 2 table:
Before performing any computations, it appears that we can partition the total variance into at
least 3 sources: 1. error (within-group) variability, 2. variability due to experimental group
membership, and 3. variability due to gender.
(Note that there is an additional source interaction - that we will discuss shortly.)
Wh a t woul d h a v e h a p p e n e d h a d w e n o t i n c l u d egednder as a factor in the study
but rather computed a simple t test? If you compute the SS ignoring the gender factor (use the
within-group means ignoring or collapsing across gender; the result is SS=10+10=20), you will
see that the resulting within-group SS is larger than it is when we include gender (use the withingroup, within-gender means to compute those SS; they will be equal to 2 in each group, thus the
combined SS-within is equal to 2+2+2+2=8).
This difference is due to the fact that the means for males are systematically lower than those for
females, and this difference in means adds variability if we ignore this factor. Controlling for error
variance increases the sensitivity (power) of a test.
This example demonstrates another principal of ANOVA that makes it preferable over simple
8/19/2012 12:08 AM
SYLLABUS
67 of 71
two-group t test studies: In ANOVA we can test each factor while controlling for all others; this is
actually the reason why ANOVA is more statistically powerful (i.e., we need fewer observations to
find a significant effect) than the simple t test.
Interaction Effects
There is another advantage of ANOVA over simple t-tests: ANOVA allows us to detect interaction
effects between variables, and, therefore, to test more complex hypotheses about reality. Let us
consider another example to illustrate this point. (The term interaction was first used by Fisher,
1926.)
Main Effects, Two-way Interaction.
Imagine that we have a sample of highly achievement-oriented students and another of
achievement avoiders. We now create two random halves in each sample, and give one half of
each sample a challenging test, the other an easy test.
We measure how hard the students work on the test. The means of this (fictitious) study are as
follows:
How can we summarize these results? Is it appropriate to conclude that (1) challenging tests
make students work harder, (2) achievement-oriented students work harder than achievementavoiders? None of these statements captures the essence of this clearly systematic pattern of
means.
The appropriate way to summarize the result would be to say that challenging tests make only
achievement-oriented students work harder, while easy tests make only achievementavoiders
work harder.
In other words, the type of achievement orientation and test difficulty interact in their effect on
effort; specifically, this is an example of a two-way interaction between achievement orientation
and test difficulty.
Note that statements 1 and 2 above describe so-called main effects.
Higher order interactions.
While the previous two-way interaction can be put into words relatively easily, higher order
interactions are increasingly difficult to verbalize. Imagine that we had included factor Gender in
the achievement study above, and we had obtained the following pattern of means:
8/19/2012 12:08 AM
SYLLABUS
68 of 71
How could we now summarize the results of our study? Graphs of means for all effects greatly
facilitate the interpretation of complex effects. The pattern shown in the table above (and in the
graph below) represents a three-way interaction between factors.
Thus we may summarize this pattern by saying that for females there is a two-way interaction
between achievement-orientation type and test difficulty: Achievement-oriented females work
harder on challenging tests than on easy tests, achievement-avoiding females work harder on
easy tests than on difficult tests. For males, this interaction is reversed. As you can see, the
description of the interaction has become much more involved.
A General way to Express Interactions.
A general way to express all interactions is to say that an effect is modified (qualified) by another
effect. Let us try this with the two-way interaction above. The main effect for test difficulty is
modified by achievement orientation.
For the three-way interaction in the previous paragraph, we may summarize that the two-way
interaction between test difficulty and achievement orientation is modified (qualified) by gender.
If we have a four-way interaction, we may say that the three-way interaction is modified by the
fourth variable, that is, that there are different types of interactions in the different levels of the
fourth variable.
As it turns out, in many areas of research five- or higher- way interactions are not that uncommon.
SPSS -OutputA
Does background music affect thinking, specifically semantic processing? To investigate this,10
participants solved anagrams while listening to three different types of background music. A
within-subjects design was used, so all participants were tested in every condition. Anagrams
8/19/2012 12:08 AM
SYLLABUS
69 of 71
were chosen of equal difficulty, and randomly paired with the different types of music. The order
of conditions was counterbalanced for the participants. To eliminate verbal interference,
instrumental Amuzak@ medleys of each music type were used, which were all 10 minutes in
length. The music was chosen to represent Classical, Easy listening, and Country styles. Chosen
for use were Beethoven, the BeeGees, and Garth Brooks. The number of anagrams solved in
each condition was recorded for analysis. Here are the data recorded for the participants:
Use 3 columns to enter the data just as they appear above. Call them bthoven beegees and
brooks
SPSS does not perform post-hoc comparisons for repeated-measures analyses.You will have to
use the formula for the Tukey test and calculate the critical difference by hand for this problem.
Remember, the general formula is:
Tukey
This gives you a critical difference (CD). Any two means which differ by this amount (the CD) or
more are significantly different from each other.
Note
MSwg = MS within groups (from ANova table output)
n = number of scores per condition
q = studentized range statistic (Table in back of textbook) H0: (mu #1 = mu #2 = mu #3) H1: (1 or
more mu values are unequal) Analyze > General Linear Model > Repeated Measures WithinSubjects factor name: music Number of levels: 3 click Add > Define
Click over variables corresponding to level 1, 2, 3 Options > Descriptive Statistics (optional, if you
want) Plots > Horizontal axis: music
8/19/2012 12:08 AM
SYLLABUS
70 of 71
Tukey
= 2.098 = 2.10; Any two groups differing by 2.10 or more are signficantly different with the Tukey
test at alpha = .05.
Summary of Tukey Results
Means from descrptives section of output: Group 1: 13.50
Group 2: 11.10
Group 3: 11.90
1 compared to 2: difference of 2.40 (greater than our CD, so this is statistically significant)
compared to 3: difference of 1.60 (less than our CD, so this is not statistically significant)
compared to 3: difference of .80 (less than our CD so this is not statistically significant
A one-way repeated-measures Anova indicated that there were significant differences in the
number of anagrams solved across the three background instrumental (muzak) conditions, F
(2,18)= 4.42, p<.027. Post-hoc Tukey comparisons indicated that more anagrams were solved
while listening to Beethoven (M=13.5) than while listening to the BeeGees (M=11.1
To summarize the discussion up to this point, the purpose of analysis of variance is to test
differences in means (for groups or variables) for statistical significance. This is accomplished by
analyzing the variance, that is, by partitioning the total variance into the component that is due to
true random error (i.e., within-group SS) and the components that are due to differences between
means. These latter variance components are then tested for statistical significance, and, if
significant, we reject the null hypothesis of no differences between means, and accept the
alternative hypothesis that the means (in the population) are different from each other.
4.5 Summary
A t-test would compare the likelihood of observing the difference in the mean number of words
recalled for each group.An ANOVA test, on the other hand, would compare the variability that we
observe between the two conditions to the variability observed within each condition. Elementary
Concepts provides a brief introduction into the basics of statistical significance testing.
4.6 Keywords
SS Error and SS Effect.: The within-group variability (SS) is usually referred to as Error variance.
This term denotes the fact that we cannot readily explain or account for it in the current design.
8/19/2012 12:08 AM
SYLLABUS
71 of 71
Dependent and Independent Variables: The variables that are measured (e.g., a test score) are
called dependent variables. The variables that are manipulated or controlled (e.g., a teaching
method or some other criterion used to divide observations into groups that are compared) are
called factors or independent variables
4.7 Questions
1. Discus Logic behind ANova.
2. What are the Applications of Anova?
4.8 References
8/19/2012 12:08 AM

Advertising 3

Uploaded by

Copyright:

Available Formats

Advertising 3

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Advertising 3

Uploaded by

Copyright:

Available Formats

SYLLABUS

http://www.vet.pctiltd.com/PCTILMS/courses/BBA 4th Semester/BBA-0...

Not for sale

KARNATAKA STATE OPEN UNIVERSITY

VIRTUAL EDUCATION TRUST

http://www.vet.pctiltd.com/PCTILMS/courses/BBA 4th Semester/BBA-0...

First Edition 2009

Copyright by Virtual Education Trust

No part of this publication may be reproduced

All product names and company names mentioned

http://www.vet.pctiltd.com/PCTILMS/courses/BBA 4th Semester/BBA-0...

Block 4 Application in Research Methodology

http://www.vet.pctiltd.com/PCTILMS/courses/BBA 4th Semester/BBA-0...

http://www.vet.pctiltd.com/PCTILMS/courses/BBA 4th Semester/BBA-0...

Unit 1 Testing Of Hypothesis-Large and Small Samples

http://www.vet.pctiltd.com/PCTILMS/courses/BBA 4th Semester/BBA-0...

1.2 Testing of Hypothesis for Proportions

http://www.vet.pctiltd.com/PCTILMS/courses/BBA 4th Semester/BBA-0...

1.3 Hypothesis tests of Differences between Means

http://www.vet.pctiltd.com/PCTILMS/courses/BBA 4th Semester/BBA-0...

1.3.1 Theoretical Basis

http://www.vet.pctiltd.com/PCTILMS/courses/BBA 4th Semester/BBA-0...

The mean of this distribution is x x = 1 1-22

The testing procedure for a hypothesis is similar to the earlier cases.

http://www.vet.pctiltd.com/PCTILMS/courses/BBA 4th Semester/BBA-0...

We then calculate the z statistic:

http://www.vet.pctiltd.com/PCTILMS/courses/BBA 4th Semester/BBA-0...

They limits of the acceptance region are : 2.58

1.4 Standard Error of the Difference between Two Proportions

The standard z statistic in this case is calculated as :

This can be best illustrated with the help of an example:

http://www.vet.pctiltd.com/PCTILMS/courses/BBA 4th Semester/BBA-0...

The overall population or hypothesized percentage of occurrence assumes that there is no

The observed z statistic is:

http://www.vet.pctiltd.com/PCTILMS/courses/BBA 4th Semester/BBA-0...

p 2: ..133 q 12: ..867 n2 : 75

therefore calculated z <critical z we accept the null hypotheses.

1.5 Uses of p- Values

http://www.vet.pctiltd.com/PCTILMS/courses/BBA 4th Semester/BBA-0...

1.6 Using the computer to Test the Hypotheses

This hypotheses was tested at a =.05

1.7 T Test for Difference between Two Sample Means

http://www.vet.pctiltd.com/PCTILMS/courses/BBA 4th Semester/BBA-0...

The underlying population is assumed to be equal for both samples.

Both use pooled stdev=10.0

1.8 Measuring the power of a Test

http://www.vet.pctiltd.com/PCTILMS/courses/BBA 4th Semester/BBA-0...

false) to be as large as possible.

http://www.vet.pctiltd.com/PCTILMS/courses/BBA 4th Semester/BBA-0...

http://www.vet.pctiltd.com/PCTILMS/courses/BBA 4th Semester/BBA-0...

illustrated as the colored area in Figure 2e.

1.9 Theoretical aspects of the t distribution

1.10 Characteristics of the t distribution

http://www.vet.pctiltd.com/PCTILMS/courses/BBA 4th Semester/BBA-0...

1.11 Degree of Freedom

http://www.vet.pctiltd.com/PCTILMS/courses/BBA 4th Semester/BBA-0...

1.12 t Values for one Tailed Tests

1.13 Hypothesis Testing Using the t Distribution

http://www.vet.pctiltd.com/PCTILMS/courses/BBA 4th Semester/BBA-0...

is the estimated standard error of the sample means.