Tell me more ×
Cross Validated is a question and answer site for statisticians, data analysts, data miners and data visualization experts. It's 100% free, no registration required.
up vote 1 down vote favorite
1

I have obtained optimally scaled variables from a highly mixed nature of data containing binary, nominal, ordinal and scale type variables. The optimal scaling was obtained in SPSS through a CATPCA procedure. Now I want to use these variables in a Factor Analysis and want to use a rotation to obtain meaningful loads. What I think is that I should avoid any normality assumption for these optimally scaled variables. So, maximum likelihood method of factor extraction is probably not applicable here (although I am not so good at FA, so not exactly sure). Should I use principal components method of factor extraction instead?

What else method can be useful that avoids distributional assumption?

If I want the factor scores in a further regression as IVs, will it be at all a good idea to use oblique rotation? I think I should try to keep the factors as uncorrelated as possible if I want them as IVs in a further regression. Is this concept right?

Thanks for any kind of suggestion. :)

share|improve this question
1  
Use principal axes or minimal residuals (better not generalized). Use PCA only if you want PCA model, not FA model. Oblique factors are never correlated greatly so that to become obstacle in futher regressions. – ttnphns Sep 14 '12 at 19:03
Thank you, does principal component method of factor extraction in Factor Analysis stand for a PCA model? Or is it a just a method of estimation of the factor model with factor loadings being the scaled coefficients of the first few sample principal components? Getting a bit confused. I was following 'Applied Multivariate Statistical Analysis', fifth edition by Richard A. Johnson and Dean W. Wichern. – Blain Waan Sep 14 '12 at 19:35
principal component method = PCA – ttnphns Sep 15 '12 at 6:01
Thank you again. – Blain Waan Sep 16 '12 at 13:34

Know someone who can answer? Share a link to this question via email, Google+, Twitter, or Facebook.

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Browse other questions tagged or ask your own question.