09-19-2011 06:15 PM
I want to use latent class approach to control for the heterogeneity in Tobit II model (sample selection model) in SAS. I suppose there is no standard package in SAS to do that and SAS IML can be use to program.
I am wondering whether anyone who can provide some hints or examples on how to used latent class approach to control for heterogeneity in Tobit model? or anyone who has an example of Tobit model estimation by using Maximum likelihood in SAS/IML (not the standard proc qlim)?
the most relevant material I found is the article " THE TOBIT MODEL: AN EXAMPLE OF MAXIMUM LIKELIHOOD ESTIMATION SAS/IML".
but they didn't introduce latent class in Tobit model.
I really appreciate any example code, comments or thoughts here. thanks a lot in advance!
09-21-2011 10:17 AM
Or, even better, the statical forum: http://communities.sas.com/community/sas_statistical_procedures
Hallahan's paper describes the Tobit model in terms of a latent variable (p. 4, second column) and shows how the output of his SAS/IML program is the same as is producced for a Tobit model created by using PROC LIFEREG. Here is a link to the LIFEREG doc that shows how to evaluate a Toibt model: http://support.sas.com/documentation/cdl/en/statug/63962/HTML/default/viewer.htm#statug_lifereg_sect...
If this doesn't answer your question, maybe you could post a reference or example that indicates what is special about your problem that can't be handled by LIFEREG.