st: [CMP]: Multinomial probit with Heckman selection

["Giacomo Zanello" <g.zanello@reading.ac.uk>]

Dear all,
I am estimating a multinomial probit with Heckman selection - at the first 
stage a probit followed by a multinomial probit. CMP (which can be obtained 
from -ssc-) seems not to be able to estimate the model. Whatever independent 
variable is used, the probit model is always collinear with one of the 
individual models estimated as starting point for the full model fit. Can 
anyone suggest a way to overcome this problem?

Following the example in the help file, this is the code I am using:

***************
preserve
webuse sysdsn3, clear
cmp setup
constraint 1 [atanhrho_23]_cons
constraint 2 [lnsig_3]_cons
cmp (insure = age male nonwhite)(site3 = age male nonwhite  ppd0),  ///
nolr ind(site3*$cmp_mprobit $cmp_probit) constraint(1 2) structural 
tech(dfp) qui
****************

Thank you very much.
Giacomo 

*
*   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/