topic Re: PCA and dichotomous variables in PROC FACTOR in SAS Procedures

PCA and dichotomous variables in PROC FACTOR

deleted_user — Wed, 20 Aug 2008 18:33:20 GMT

I'm working on doing a principle component analysis and using the PROC FACTOR to do this. My issue is that I have all dichotomous variables. How does PROC FACTOR handle this?

Would it be better to first calculate the tetrachoric correlations in PROC FREQ?

Thanks.

Re: PCA and dichotomous variables in PROC FACTOR

deleted_user — Tue, 26 Aug 2008 20:09:42 GMT

if no one can answer the question, could someone at least tell me other places I could look for the answer?

thanks !!

Re: PCA and dichotomous variables in PROC FACTOR

Cynthia_sas — Tue, 26 Aug 2008 22:12:01 GMT

Hi:
In the SAS/Stat manual, there is a discussion of PROC PRINCOMP (principal component analysis) versus PROC FACTOR:
http://support.sas.com/documentation/cdl/en/statug/59654/HTML/default/statug_intromult_sect002.htm

Overview on Multivariate Procedures:
http://support.sas.com/documentation/cdl/en/statug/59654/HTML/default/statug_intromult_sect001.htm

Tech Support note:
http://support.sas.com/kb/22/558.html

For immediate or urgent help, you might consider opening a track with Tech Support.

cynthia

Re: PCA and dichotomous variables in PROC FACTOR

deleted_user — Thu, 28 Aug 2008 15:07:41 GMT

This is really more of an academic question than a technical one. I've used PROC FACTOR with dichotomous variables, but the question is, should it be done? I tried to find research on this topic a few years ago, but didn't have any success.

Re: PCA and dichotomous variables in PROC FACTOR

Paige — Thu, 28 Aug 2008 16:52:09 GMT

There's no way to answer "should it be done" unless you tell us more about what your data is, and what you are hoping to learn from your data. Usually, I see no point in performing PCA on dichotomous variables. But unless I know more about what you are trying to accomplish, I can't point you in any particular direction.

Re: PCA and dichotomous variables in PROC FACTOR

AA1973 — Fri, 27 Apr 2012 01:42:22 GMT

see thread

Re: PCA and dichotomous variables in PROC FACTOR

Ksharp — Fri, 27 Apr 2012 04:24:19 GMT

It looks like your data is catagory data .How about using Correspondence Analysis.