st: RE: Estimation selection model with selmlog/svyselmlog using sample weights

[Nick Cox <n.j.cox@durham.ac.uk>]

-bsample- has long been an official command, so I wouldn't expect to find a detailed digression on its use within the help for a user-written command. I don't know how to apply it to your problem. 

The authors of the other -*selmlog- commands are probably not members of Statalist. Note that they didn't respond the first time you asked this. 

Nick 
n.j.cox@durham.ac.uk 

David Wiesinger

i'm apologising for being so imprecise, it's my first time that i'm posting in this forum.

First i was estimating my selection model using the original "selmlog" (written by Bourguignon et al) command and i got a result. Unfortunatly i wasn't able to include my probability weights.

For that reason, i was trying to use "svyselmlog" ( i got it form the homepage http://fmwww.bc.edu/RePEc/bocode/s and it was written by R. E. De Hoyos). 

with 
version 8: svyset [pweight=Gewicht]  
svyselmlog continuous choice, select( discrete choice), meth(dmf2) bootstrap(20) force

i got the error message

"PSU is equal to the number of observations, svybsamp2 is not necessary. Use -bsample-"

Unfortunatly i don't know, how to use -bsample-, becasue i couldn't find a hint in the help-file for svyselmlog.

Nevertheless i wanted to try out the procedure, so i created 2 random psu's and i rerun the model

version 8: svyset [pweight=Gewicht], psu(random_psu)
svyselmlog continuous choice, select( discrete choice), meth(dmf2) bootstrap(20) force

and i got an "conformability error" after some bootstrap steps. When the error occures, seems to be dependent on how many variables i put into the selection term and the used procedure. The less variables, the more steps can be calculated.
when i am estimating the first stage with svymlogit, the model converges and i get a result.
When i'm estimating the same model with selmlog, it works.

It is very likly, that i was making a stupid mistake, but for now i can't find it. How can i estimate the model with my weights only? 

*
*   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/