Statalist


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

st: RE: confirmatory factor analysis with binary variables


From   "Verkuilen, Jay" <JVerkuilen@gc.cuny.edu>
To   <statalist@hsphsun2.harvard.edu>
Subject   st: RE: confirmatory factor analysis with binary variables
Date   Mon, 15 Dec 2008 15:45:51 -0500

>>I am trying to do a confirmatory factor analysis on data that is all
binary, 0=no, 1=yes.   I have downloaded the tertrachoric command and
used this to find the tertrachoric correlations.  Is it possible to do
confirmatory factor analysis with this data using the cfa1 command in
Stata 9? If so, can anyone explain how?<<

A really cheesy workaround might be to do the following:

(1) Get tetrachoric correlations using -tetrachoric-.
(2) Simulate a dataset of the proper sample size with that correlation
matrix. 
(3) Using -factor- with ML extraction and one factor, or with -cfa1-,
fit a one-dimensional factor analysis. 

This is what my SEM professor Rod McDonald would have called a "device"
but it might give you a very approximate answer until you get -gllamm-
or MX up and running and do it right. 

To skip the whole simulation bit, use -factormat- with the tetrachoric
correlations via ML factor analysis, which is basically a one-step
version that got generalized and fixed up into the algorithms in
software like Mplus or MX. People used to do this back in the old
days---you'll see it discussed in older factor analysis books. Oblique
target rotations can even approximate a CFA tolerably well. 

JV

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/



© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index