Essentials of
Research
Day 1
Dr Shweta
Pandey
What will you learn in this session
Uses of factor analysis
Conditions for factor analysis
Factor rotation
Interpretation of factors
Factor analysis
1. Multi-variate statistical technique
2. No distinction in dependent and independent variable
3. It allows researchers to investigate concepts that are not easily measured directly
by collapsing a large number of variables into a few interpretable underlying
factors.
3
Factor analysis
1. Data reduction method: Extraction of factors where factor is a linear
combination of variables.
2. Multiple observed variables have similar patterns of responses because they
are all associated with a latent (i.e. not directly measured) variable. For
example, people may respond similarly to questions about income,
education, and occupation, which are all associated with the latent variable
socioeconomic status
4
Conditions for Factor Analysis
• Continuous data: Interval (Likert)/ Ratio scale
• Standardization of scales:
(Actual score- mean score of the statement)/ Standard deviation of the statement
• Initial set of variables should be highly correlated
Principal Component Analysis
Factors are linear combinations of the variables which are supposed to be
highly correlated, the mathematical form of the same could be written as
• Principle component analysis chooses those values of W which ensure
that first factor explains the maximum portion of the variance
Conditions for Factor Analysis
• Barlett’s Test of Sphericity: A test statistic used to examine the
hypothesis that the variables are uncorrelated in the population.
• H0: Correlation matrix is insignificant (the population correlation
matrix is an identity matrix; each variable correlates perfectly with
itself (r = 1) but has no correlation with the other variables (r = 0))
• H1: Correlation matrix is significant
Statistics associated with factor analysis
• Kaiser-Meyer-Olkin (KMO) Measure : An index used to examine
the appropriateness of factor analysis.
• Compares magnitudes of observed correlation coefficients with
magnitudes of partial correlation coefficients.
• High values (between .5 and 1.0) indicate factor analysis is
appropriate.
• Values below .5 imply that factor analysis may not be appropriate.
Conducting factor analysis
Formulate the problem
Construct the correlation/ covariance matrix
Determine the method of factor analysis: orthogonal/oblique/varimax
Determine the number of factors
Rotate the factors: Varimax etc.
Interpret the factors
Using factor scores
Factor analysis in SPSS
Click on Analyze and
then select Dimension
reduction and then
factor
Factor analysis in SPSS
Select all variables and
then click on the arrow
pointing right
Factor analysis in SPSS
After clicking on the
arrow all variables will
show in the Variables
section
Next Click on
Descriptive tab and
select Univariate, Initial
solution and KMO and
Bartlett’s test
Factor analysis in SPSS
Next Click Extraction
and choose Principal
component and click
on Continue
Choose rotation method
Select ROTATION and
select the specific
method of rotation:
Varimax
Cleaning up the result
Select Options and
suppress small
coefficients. Fill value
Saving factor scores as variables
Select Scores and
click on save as
variables
Cleaning up the result
Select Options and
click on sorted by size
Factor loadings
Variable Factor 1 Factor 2
Factor loading of 0.793 is the
Easy accessibility 0.793 0.047
correlation between the
Maturity time 0.765 -0.198 variable Easy accessibility and
Component (Factor) 1 or factor
Returns 0.527 0.160
score 1
Tax rebate 0.309 0.125
Risk awareness -0.176 0.753
Insurance Cover 0.335 -0.707
Credibility 0.570 0.633
Eigen value 2.054 1.551 Eigen value of Factor
1= 0.793^2+
0.765^2+….+0.570^2
Factor= f( Items/ Variables)
Factor loadings: Correlation of each item/variable with the Factor score
Eigen values
Only factors with Eigen value of at least 1 are accepted
28.78%
variance is
explained by
Eigen factor 1
value>1
Communalities
• How much of the
item/variable is
explained by the
Factors
For example a
community of 0.598
shows that 59.8% of
the variable/item is
explained by the
factors
Rotation of factors
• To make the solution easy to interpret we do rotation of initial factor solution
• There are options for orthogonal rotation, varimax rotation, oblique rotation
• The total variance explained by all factors taken together remains the same
after rotation.
• The communality's also remain the same
• Factor loadings may change
Extraction of factors
• Factor= f( Items/ Variables)
• Factor loadings: Correlation of each item/variable with the factor
• Factor Scores: Is computed for each of the factors corresponding
to each respondent. Factor scores are statistically independent and
can be used as independent variables
Statistics associated with factor analysis
• Factor Matrix: Contains the factor loadings of all the variables on all the factors
extracted.
• Percentage of Variance: The percentage of the total variance attributed to each
factor.
• Communality: This is the proportion of variance explained by the common
factors. The amount of variance a variable shares with all the other variables
being considered.
