Path: Janda's Home Page > Workshop > Factor Analysis > Outline

In the unrotated factor structure, four variables load high on unrotated Factor I, but two load in opposite directions. Factor II clearly reflects feelings toward Perot, but Factor III is undefined.

Rotation usually makes a factor structure more interpretable. The rotated analysis invites us to name a "Republican," a "Democratic," and a "Perot" factor to describe the feeling thermometer data.

Unrotated factor analysis
Rotated Analysis
Repub.
Democrt
Perot
0.89
-0.23
-0.02
0.84
-0.24
-0.00
-0.27
0.87
0.01
-0.31
0.79
0.06
0.03
0.02
0.87
-0.02
0.06
0.87
0.85
-0.19
0.02
-0.12
0.83
0.03

Factor I
Factor II
Factor III
Bush: Pre-Election
-0.81
0.13
0.42
Bush: Post-Election
-0.78
0.14
0.37
Clinton: Pre-Election
0.79
0.07
0.45
Clinton: Post-Election
0.77
0.10
0.36
Perot: Pre-Election
0.05
0.85
-0.18
Perot: Post-Election
0.11
0.85
-0.18
Quayle: Pre-Election
-0.75
0.17
0.41
Gore: Pre-Election
0.65
0.11
0.52

There are various methods for rotating factors. The first choice is between orthogonal rotation (which keeps the factors uncorrelated) and oblique rotation (which allows the factors to be correlated). Orthogonal (used here) is the most common.

The next choice deals with the way the loadings are evaluated. The most common method (used here) is varimax, which maximizes the variance of the squared factor loadings.

End of workshop. Return to the lecture outline.