Not an expert, but I know that -mtreatnb- (sj6-4) fits a treatment-effects model that considers the effects of an endogenously chosen multinomial treatment on another endogenous count outcome, conditional on two sets of independent variables. The treatment variable is modeled via a multinomial logistic and the outcome via a negative binomial regression. The model is fitted using maximum simulated likelihood.
Maybe you can find something from that way. Or ask privately to the authors if they have some suggestions
Nicola
P.S. Direct e-mails to my address will be filtered. Offline until 17th August
At 02.33 29/07/2008 -0400, "James Shaw" wrote:
>Dear Statalist,
>
>Is anyone aware of an article discussing (two-stage) selection
>modeling in the context of multinomial/multivariate probit? I have a
>selection indicator that is polytomous and would prefer to treat it as
>multinomial when fitting the model. While -mvprobit- or -mprobit- can
>be used to estimate the coefficients of the first-stage selection
>equations, it is not clear to me how one would derive the inverse
>Mills ratio (or its analog) or how this would enter into each of the
>second-stage outcome equations.
>
>I have scoured the literature and have found nothing relevant.
>Cappellari and Jenkins have published on the use of multivariate
>probit models to account for selectivity when the outcome is binary.
>Also, I am aware of Stata's -selmlog- command and the related
>literature. However, -selmlog- appears to require that the outcome be
>missing for all save one of the selection categories. This is not the
>case in my situation.
>
>Thank you in advance.
>
>Regards,
>
>Jim
*
* 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/
