Page 210 - Quantitative Data Analysis

Quantitative Data Analysis
Simply Explained Using SPSS

In most cases, PCA and EFA will produce similar results in terms of
data structure and interpretations of “factors” and “components”.
Following are the some similarities in PCA and EFA discussed by
Diana D. Suhr in her paper titled “Principal Component Analysis vs.
Exploratory Factor Analysis”.

1. PCA and EFA both are variable reduction techniques. The
combines set of variables into component (s) or factor(s).

2. PCA and EFA have common assumptions including the
normality of each variable.

3. Scale of measurement for variables is interval or ratio type

4. Random sampling – at least 5 observations per variable and
at least 100 observations.

5. Large sample size is recommended for more stable
estimation. Specially using maximum likelihood estimation,
10-20 observations per variables is recommended to
compensate for missing cells (missing one or more variable
values).

6. EFA and PCA both identify relationship between observed
variables.

7. In PCA and EFA each pair of observed variables has bivariate
normal distributions.

8. If communality of PCA and EFA are large enough (close to
1), results could be identical.

In factor analysis or principal components analysis, having items
load together as hypothesized does not provide direct evidence that
your items are measuring the construct that you defined in your
operational definition. However, if the items do function
mathematically as you hypothesized, it provides evidence to

The Theory and Applications of Statistical Inferences 194