Page 213 - Quantitative Data Analysis

Quantitative Data Analysis
Simply Explained Using SPSS

EFA includes error term in the
PCA assumes perfect reliability
6 of each observed variables (no equation which means EFA
includes a unique variance
error variance)
(residual variance).
PCA has initial communality of 1 EFA does not have initial
7
for each variable communality of 1.
EFA is used to combine the set
of correlated variables into one
or more factors (latent variable
PCA is used to combine the set
/unobservable variables) to
8 of correlated variables into one provide an operational
or more components (variables)
definition for an underlying
construct by using that
observed variables.
There is no problem of extreme Extreme collinearity in EFA will
9
collinearity in PCA mess up the results.
In PCA, the correlation matrix In EFA, the correlation matrix
10
contains 1 on diagonals. contains unique factor

Technically, a PCA is a linear combination of the variables that
accounts for the maximum variance: The first component extracted
in PCA accounts for a maximum amount of total variance in the
observed variables, the second uncorrelated component accounts
for next largest variance and this process goes so on. Hence,
principal component analysis is the procedure to use the set of
correlated variables and transformed into set of uncorrelated
components or variables (Stevens, 2012). The components are
interpreted by using the components -variable correlations that is
called factor loadings (Stevens, 2012). Factor analysis using
maximum likelihood (ML) estimation requires large sample size and

it is one of the drawbacks of EFA.

The Theory and Applications of Statistical Inferences 197