I have some questions with respect to Stata's factor analysis method pcf:
Telling from my experience pcf always (!) yields clearly 'better' solutions
than the other factor methods pf, ipf, and ml (e.g. higher factor loadings,
lower uniqueness, and sometimes a more stable factor structure). Any ideas
why is that? Would you interpret pcf results in the same way you interpret
results of other methods? Or, putting it another way, is pcf in this sense
superior to the other methods?
The Reference Guide states that the pcf method assumes the communalities to
be 1. In my understanding this is a quality of principal components
analysis. Is pcf a method for principal components and not for principal
factors? Puzzling to me is that in any analysis I've done pcf finds its own
quite heroic assumption to be wrong, i.e. the items do have some unique
variance. Thus it seems pcf results contradict pcf assumptions. What am I
missing here?
Finally, I'd like to use the calculated factor scores in further regression
analysis. Telling from the figures, with pcf the latent constructs in
question are measured in a 'better' way than with the other methods. One
could argue that the pcf solution yields a clearer structure, accumulates
more variance on the common factor, and thus helps to measure construct Ksi
better. Would you follow? Is it advisable to use pcf if one wants to obtain
factor scores?
Any help and thoughts appreciated!
Best, Kai
_____________________________________
Kai Muehleck
International Social Justice Project (ISJP)
Institut fuer Sozialwissenschaften
Humboldt-Universitaet zu Berlin
Unter den Linden 6
10099 Berlin
Germany
+49-30-2093-5382
+49-30-2093-4430 (Fax)
[email protected]
www.isjp.de
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
