Factor analysis

MBA-BM (2024-26, Sec A&B)

Prof. Nirali Shah

E-mail: nirali@xim.edu.in
 Factor analysis

• Primarily used for data reduction & summarization

• Reduces data complexity

• To manageable levels without loss of information
• Combines highly correlated variables into a single factor

2
 Factor analysis

• Used in the circumstances to identify

• An underlying dimension (factor) to explain the correlation

among a set of variables

• A new smaller set of uncorrelated variables to replace the

original set of correlated variables in the subsequent multivariate
analysis

• A smaller set of variables from a larger set for use in subsequent

multivariate analysis

3
 Factor analysis: Applications in marketing

• Marketing segmentation to identify underlying variables

• Product research to identify important attributes & benefits

• Promotion studies to understand media consumption habits

• Pricing studies to identify characteristics of consumers

4
 Statistics associated with factor analysis

• Bartlett’s test of sphericity

A test statistics used to examine the hypothesis that the variables
are uncorrelated in the population
Each variable correlates perfectly with itself (r=1) but has no
correlation with the other variables (r=0)

• Correlation matrix:
Displays the correlation of each variable with every other variable

5
 Statistics associated with factor analysis

• Communality
The percentage of variance explained by all the factors extracted

• Eigenvalue
The amount of variance explained by an extracted factor

6
 Statistics associated with factor analysis

• Factor loading
The correlation of a variable with a factor
Used to interpret factors

• Factor matrix
Exhibits the factor loading of all the standardized variables on all
the factors extracted

• Factor scores
Values for each respondent on each extracted factor

7
 Statistics associated with factor analysis

• Residuals
The difference between the original correlations exhibited in
the correlation matrix and the factor loading as depicted by
the factor matrix

• Percentage of variance
This is the percentage of total variance attributed to each factor

8
 Statistics associated with factor analysis

• Kiser-Meyer-Olkin measure of sampling adequacy

To examine the appropriateness of factor analysis
Values above 0.5 indicate that factor analysis is appropriate
Values below 0.5 imply that factor analysis may not be appropriate

• Scree plot
A graph of the eigenvalues plotted against the number of
factors in order of extraction

9
 Statistics associated with factor analysis
Eigenvalue

Scree Plot

1 2 3 4 5
Component/Factor Number 10
 Conducting factor analysis
Formulate the problem

Construct the correlation matrix

Determine the method of factor analysis

Determine the number of factors

Rotate the factors

Interpret the factors

Calculate the Select the

factor scores surrogate variables

Determine the model fit 11

 Formulate the problem

• Objectives of factor analysis

• Variables
• Identify variables to include
• Use past research, theory, and judgment of researcher
• Appropriately measure variable – interval or ratio scale

• Appropriate sample size

• Rule of thumb: Sample size at least 4 to 5 times variables

12
 Construct the correlation matrix

• For factor analysis to be appropriate: The variables must be correlated

• Use of Bartlett's test of sphericity

• Null hypothesis: The variables are uncorrelated in the population

• Use of Kaiser-Meyer-Olkin (KMO) measure of sampling adequacy

• A value greater than 0.5 desired
• Small values (smaller than 0.5) indicate
• The correlations between pairs of variables cannot be explained by
other variables
• Factor analysis may not be appropriate
13
 Determine the method of factor analysis

• Method for factor extraction

• Most popular method: Principal component analysis

• Objective
• To determine minimum number of factors that will account for
maximum variance

14
 Determine the number of factors

• A priori determination
• Drop in variance
• Computation of an Eigen value
• Percentage of variance criterion
• Significance test criterion
• Scree plot criterion
• Split-half reliability

15
 Determine the number of factors

• Drop in variance Factor % of variance

• Look for the point at which 1 45.520
there is a sudden and
2 36.969
substantial drop in the
variance explained 3 7.360
between two points
4 5.688
5 3.044
6 1.420

16
 Determine the number of factors

• Computation of an Eigen value Factor Eigen value

• Higher the Eigen value of a factor: Explains 1 2.731
higher amount of variance
• Assumption before extraction: Original 2 2.218
variable has an eigenvalue = 1
• After extraction: Retain factors having an 3 0.442
Eigen value of more than 1 4 0.341
5 0.183
• Significance test criterion 6 0.085
• Test the statistical significance of each
eigenvalue
• Retain only those factors with a significant
eigenvalue
17
 Determine the number of factors

• Percentage of variance criterion Factor % of Cumulative%

variance
• Cumulative variance explained 1 45.520 45.520
by the factors at various levels 2 36.969 82.488
• When cumulative variance 3 7.360 89.848
reaches satisfactory level, no
further factors need be 4 5.688 95.536
extracted 5 3.044 98.580
• ‘Satisfactory’ level: Subjective 6 1.420 100.000

18
 Determine the number of factors

• Scree plot criterion

• Graph of eigenvalues
against the number of
factors extracted

Eigenvalue
• Point at which there is a
steep drop in the graph is
indicative of the numbers of
factors should be extracted

1 2 3 4 5
Component Number

19
 Determine the number of factors

• Split-half reliability
• Splitting the sample randomly

• Perform factor analysis on each half

• Retain factors with high correspondence factor loadings across

the two subsamples

20
 Rotate the factors

• Each factor to have nonzero, or significant, loadings or

coefficients for only some of the variables.

• Each variable to have nonzero or significant loadings with

only a few factors, if possible, with only one.

21
 Rotate the factors

• The rotation is called orthogonal rotation if the axes are

maintained at right angles.

• Preferred method of orthogonal rotation : Varimax procedure

• Varimax procedure
• Minimizes the number of variables with high loadings on a factor
• Orthogonal rotation results in factors that are uncorrelated

22
 Rotate the factors

• Use factor matrix:

• Identifying which factors associated with which original variables

• Loading of each variable on each of the extracted factors

• Loading similar to a correlation matrix
• Loading values between 0 and 1
• Value near to 1 present high loadings
• Value near to 0 present low loadings

• Objective:
• To find variables which have a high loading on one factor,
but low loadings on other factors

23
 Interpret the factors

• Identifying the variables that have large loadings on the same factor

24