In order to develop a parsimonious representation for
the various constructs in the survey, we initially conducted
a principal-components-based exploratory factor analysis
for each set of questions from the individual sections of
the survey instrument.
16
We eliminated the few elements
that had a cross-loading greater than 0.50 or that loaded
onto a factor that did not make logical (or theoretical) sense.
After all of the constructs were defined, we performed a second factor analysis to verify the initial results. We included
only those elements from the results of the initial factor
analysis related to the six constructs used in this study.
Using the principal components method, the same six constructs emerged with eigenvalues greater than 1.0, accounting for 62% of the total variance in the data. These factors
were in general alignment witha prioriexpectations. The
oblique
17
rotation resulted in the following factors: