> -----Original Message-----
> From: [email protected] [mailto:owner-
> [email protected]] On Behalf Of Dan Rigby
> Sent: Friday, January 28, 2005 9:13 AM
> To: [email protected]
> Subject: st: Regression to explain simultaneously determined probabilities
>
>
> I have estimated a latent class model with 3 classes for 334
> individuals. For each individual in the sample I have estimates of
> their (3) class membership probabilities.
>
> I would like to regress these class membership probabilities on
> demographic variables, but need to take account of the fact that (i)
> the probabilities are simultaneously determined and sum to 1 for an
> individual, and (ii) each probability is bounded by 0 and 1.
>
> Sureg will predict out of the feasible range so does anyone have a
> suggestion on the best way, in Stata, to undertake such a
> regression.
>
> thanks
> dan rigby
>
Dan,
How about transforming the dependent variable ln(y/1-y) with -sureg- or
using using -glm, family(binomial) link(logit)- with -suest-. You might
find the FAQ "How does one estimate a model when the dependent variable is a
proportion?" at http://www.stata.com/support/faqs/stat/logit.html useful.
Hope this helps,
Scott
*
* 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/
