Professional Documents
Culture Documents
Subject PSYCHOLOGY
Module No and Title Module No. 12: Factorial Design: 2- Way Analysis of
Variance
Module Tag PSY_P2_M12
TABLE OF CONTENTS
1. Learning Outcomes
2. Introduction
3. Analysis of Variance
3.1 What is ANOVA
3.2 What are Factorial Designs
3.3 What is Two-Way ANOVA
3.4 Types of Two-Way ANOVA
4. Assumptions, Models and procedure of Two-Way ANOVA
4.1 Assumptions of Two-way ANOVA
4.2 Models of two – way ANOVA
4.3Procedure of Two-way ANOVA
4.4 Advantages of Two-way ANOVA
4.5 Applications of two- way ANOVA
5. Numerical application of ANOVA
6. Using IBM SPSS 20 for two –way ANOVA
7. Summary
1. Learning Outcomes
After studying this module, you shall be able to…
2. Introduction
With changing times, people have become more conscious of their physical health and
appearance. Various exercise patterns have come up to lose weight and stay fit, whether it be our
childhood beloved rope skipping or six pack building pushups or the much more exciting and
enjoyable aerobics. But how do we choose which exercise is effective for losing weight quickly.
One can find that out in a scientific way by comparing the weights of the individuals doing each
of these exercises after a specific period of time. We might take 5 females doing each of these
exercises and compare their weights after 4 weeks to find the effect of the exercises. Here, we
bring ANOVA to compare the 3 groups. After the suitable statistical analysis the researcher
would come to know if there is a significant difference between the various exercise patterns and
suitable applications of the results can be made.
3. Analysis of Variance
The statistical model used to compare more than two conditions is called ANOVA or analysis of
variance. The readers must know that t-test is used to compare two means only. This is because if
one was interested in comparing 3 experimental conditions and if one was to carry out t-tests on
each pair of groups, then one would have to conduct three separate tests, one comparing group 1
& 2, other comparing 2 & 3 and the last comparing 1 & 3. One would also recall the concept of
type I and type II errors, with each statistical test the probability of occurrence of these errors
increases too. This can be explained as:
At a .05 level of significance, the type I error or the probability of falsely rejecting the null
hypothesis is only 5%. Thus, the probability of no Type I error is .95 i.e. 95% for each test. If it is
assumed that each test is independent, one can multiply the probabilities giving an overall
probability of no Type I error as .95 x .95 x .95= .857 as the probability of no type I error is .95
for each test and there are 3 tests. The probability of making at least one type I error becomes 1 -
.857 = .143 i.e. 14.3 increasing the probability of making Type I error from 5% to 14.3%,
unacceptable to social scientists.
So, it is clear that with increasing number of comparisons the probability of errors increases too.
As the number of groups for comparison will increase, the number of t-tests to be carried out will
also increase increasing the error rate thereafter. Therefore, ANOVA is brought to rescue. And
this is the one way ANOVA discussed in the earlier modules. One way ANOVA compares three
or more levels of one factor. With increased number of factors of variables the research design
becomes more complex. The researchers resort to factorial designs and higher order statistical
procedures too. This paves the way for ANOVA for factorial designs (2 way ANOVA).
When an experiment has two or more independent variables, it is known as a factorial design with
the variables being referred to as factors. Factorial designs can be:
(the readers may refer to modules number 3& 4 for a detailed discussion of research designs)
When one uses ANOVA to analyze a situation in which there are two or more independent
variables, it becomes factorial ANOVA. The Two Way ANOVA extends one way ANOVA to
include two independent or explanatory variables. It is like an experimental study that gives
numerical results with two explanatory variables which are categorical in nature. The independent
variables are referred to as factors and may have further levels. It is assumed that the dependent
variable is affected by two factors. A subject may be exposed to a level of each of the two
explanatory categorical variables. As there are two independent variables involved the impact that
change in an independent variable may produce on the outcome may not be related to the other
independent variable. Let us explain this with the help of an example
A company wishes to understand the occupational stress in response to age and gender, the
dependent variable becomes occupational stress. Age and Gender become the two independent
variables. If the employees are categorized according to age groups into age less than 40, 40-55
and above 55 years in addition to the gender classification being male and female, the design
requires the use of Two-way ANOVA. Here, the factor age has 3 levels and the factor gender has
2 levels. This implies, there are 3 x 2 = 6 groups or cells. And on this layout, one obtains the
occupational stress scores from the employees categorized in the 6 cells.
Two way repeated measures ANOVA-two independent variables both measured using
the same participants
Two way mixed ANOVA- two independent variables, one measured using different
participants and the other measured using the same participants
The Two-way ANOVA as an analytic method is appropriate for a study with a quantitative
outcome and two or more categorical explanatory variables,
Normality: it is assumed that the parent populations from which the samples have
been drawn are normally distributed
Independence: the chosen samples are independent of each other.
Homogeneity of Variances: it is assumed that the scatter of scores or the variance
among the samples is uniform or in other words the obtained scores are uniformly
scattered around their means in each sample.
Sample size: it is important that the samples chosen are of the same size.
Structural Model with interaction: the structural model of ANOVA suggests that the
combination of explanatory variables being explored have their respective populations
with corresponding means. The interactive patterns of these means may follow any
arbitrary pattern. This would imply that the outcome produced by changing any one
variable may depend upon the levels of other variable.
To begin with, Two-way ANOVA requires formulation of hypotheses which are: the
first factor population means are equal; the second factor population means are equal and
the two factors do not show any interaction.
The treatment groups are then formed by making all possible combinations of the two
factors. For instance, if the first factor has 3 levels and the second one has 2 levels, there
will be 3 x 2 = 6 different treatment groups.
For instance, one can assume the treatment of a particular psychological disorder. The
type of psychotherapy and the medication being used for treatment of that disorder are
taken into consideration. There are 5 types of fertilizers that one has and 3 types of
medication one is going to use. Therefore, this group will have 15 treatment groups. So
the degrees of freedom for the groups would be: type of medication = 2 (3-1); type of
psychotherapy = 4 (5-1); degrees of freedom for interaction between medication and
psychotherapy = 2 X 4 = 8. The exhibit given below describes the given arrangement of
the data.
Medicine 104, 108 94, 101 93,106 102, 103 99, 101
A
Medicine 109, 111 97, 98 99, 100 107, 111 104, 106
B
The main effect will involve the independent variables one by one, taking into account only the
rows and columns and not mixed similar to ANOVA. The interaction effect is the effect one
factor has on the other factor, degrees of freedom being the product of the two degrees of freedom
for each factor. Within variation is the sum of squares within each treatment group i.e. one less
than the sample size for each treatment group. The total number of treatment groups is the
product of the number of levels for each factor. Within variance is the within variation divided by
its degrees of freedom.
F test is run for each of the hypotheses i.e the mean square for each main effect and interaction
effect divided by within variance. The 2 way ANOVA table gets represented as follows:
Source SS df MS F
Sum of
Total N-1, abn- 1
others
The 2- way ANOVA is applied to the data when the researcher wishes:
Focus of analysis: the purpose of the 2-way or multi factorial ANOVA is to analyze the
mean differences across main effects, interaction effects and simple effects.
Number of independent variables: the independent variable is referred to as a factor here
so, it is explanatory enough that more than one independent variable is involved in multi
factor ANOVA.
Scales of measurement of dependent variable: the dependent variable should be
continuously scaled so, it should either be in interval or ratio scale.
Relationship of the participants across groups being compared: the multi factor ANOVA
may be used for within group (related subjects) or between group (independent) designs.
The results of the two way ANOVA indicated a significant main effect for gender F(1, 42) =
22.37, p < .001and a significant main effect for the length of exercise program, F (2, 42) = 36.03,
p < .001. the results in addition also showed a significant interaction between gender and length
of exercise program, F (2, 42) = 11.84, p < .001 indicating that any differences between the
length of exercise programs were dependent upon which gender the subjects were and that any
differences between females and males were dependent upon which length of exercise program
they were in. Approximately, there were 14% of the total variance of the flexibility levels was
attributed to the interaction of gender and length of exercise program.
Source SS df MS F p
Gender 330.75 1 330.75 22.37 0.000
Duration of exercise 1065.50 2 532.75 36.03 0.000
schedule
Gender x Duration 350.00 2 175.00 11.84 0.000
of exercise schedule
Within (error) 621.00 42 14.79
Total 2367.25 47
The results of ANOVA show that the interaction between duration of exercise schedule and the
gender is significant. Let us study the gender effects first i.e. the difference between the males
and females on the threelevels of the duration of exercise schedule. The significant differences
have been reported only on the 3 weeks duration schedule. Females had significantly higher
levels of flexibility also.
One also studied the duration ofexercise schedule main effects and the results are there in the
table. One can check out that the females in the 3 weeks duration had significantly higher
flexibility than the ones in 1 week duration. Similarly for the males the results showed 1 week
duration to be significantly different from the 3 weeks duration on flexibility.
1. Treatment condition ( with two levels: standard care and standard care + contingency
management) and
2. Treatment status (number of prior treatment attempts)
The 2 x 2 factorial research design has been used. The obtained data were used to prepare
the data sheet and the following SPSS commands may be used to get the 2- way ANOVA
results:
Click Analyze>click General Linear model> click Univariate > click over Treatment
Retention under Dependent Variable and treatment Condition and Treatment Status
under Fixed Factors.
Click on the Plots button> click treatment Condition to Horizontal Axis and Treatment
Status to Separate Lines> click the Add button> click Continue.
Click on the Options button>click on the boxes of Descritive Statistics, Estimate of
Effect Size, Observed Power, and Homogeneity tests> the SIGNIFICANCE level
should be .01. click on CONTINUE. Click OK.
Save generated output as Two – way ANOVA results
7. Summary
Two- way ANOVA technique is used in case of the data being classified on the basis of
two factors. It may have repeated measurements of each factor or may not have repeated
values. It has extreme utility in various fields like social sciences, economics, biology,
business and other disciplines.
Where the z test and t test judge the significance of difference between means of two
samples, ANOVA helps perform simultaneous test on more than two sample means at the
same time. Using tow-way ANOVA, one can study how various factors interact to
produce a certain outcome
Two way ANOVA or the factorial design ANOVA follows additive and
interaction models to understand the effect of independent variables on the
dependent variable
It has widespread applicability in areas of social, industrial researches since the
most of the phenomena in today’s world are not affected by one to one linear
relationship.