0
$\begingroup$

When developing an instrument involving ordinal data (likert scale with 5-6 response levels), how does one reduce the initial item pool before completing the exploratory factor analysis? I have seen that categorical PCA (in SPSS) has been recommended here but what are additional options that are available outside of SPSS?

$\endgroup$
  • $\begingroup$ reduce the initial item pool before completing the exploratory factor analysis? That sounds a bit strange. Although there exist statistical approaches to drop some "bad" items before FA, the main tool to select good items while developing/validating a questionnaire via FA is FA itself. It is typically done and redone several times in the process of selection of the items from their initial pool. If one is planning to do nonlinear FA at that, that is, CATPCA-then-FA (instead of just FA), nothing basically changes: for every set of items he's testing he first does CATPCA then FA. $\endgroup$ – ttnphns May 30 '16 at 23:01
  • $\begingroup$ I probably did not word things appropriately, but my understanding is that PCA is first used as a data reduction technique before determining the underlying factor structure and further refining the item pool using FA (EFA and parallel analysis) (e.g. Matsunaga, 2010 How to Factor Analyze Your Data Right). $\endgroup$ – user116948 May 30 '16 at 23:20

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.