(Hopefully no stupid Outlook file attached. Apologies if so---it is a bad penny....)
Here are some options:
-As mentioned, factor analysis or PCA of polychoric correlations, but this can be... ugly, as polychoric correlations can be non-PSD.
-GLLAMM can set up categorical factor analysis by maximum likelihood.
-For "true" categorical PCA (factor analysis and PCA are quite different models because FA explicitly accounts for error whereas PCA does not), look at multiple correspodence analysis, which Stata 10 has. Joint correspondence analysis (which Stata 10 also does) is the categorical variable analog of factor analysis of a certain sort---I think it is the nominal variable analog of MINRES but don't quote me on that.
Reference for CA: Classic is Michael Greenacre's 1984 book, but it is old; he has an updated basic book. JC Gower also wrote a good one in the CRC series. Gifi is a classic as well but difficult.
Jay
-----Original Message-----
From: [email protected] on behalf of Philipp Rehm
Sent: Thu 2/28/2008 6:44 PM
To: [email protected]
Subject: Re: st: Stata module similar to PCA but for categorial variables
.
PCA = Principal components analysis?
If so, this may be useful:
-factormat- (Stata 9 or higher, I believe)
net describe polychoric, from(http://web.missouri.edu/~kolenikovs/stata)
http://www.gseis.ucla.edu/courses/ed231a1/notes2/morefa.html
help tetrachoric
HTH,
Philipp
Luhang Wang wrote:
> Dear all,
>
> I've been wondering whether Stata can do PCA alike analysis with
> categorical variables. Any suggestions? Thanks.
>
> Best,
> Luhang
> *
> * 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/
>
*
* 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/
<<winmail.dat>>
