Re: st: FW: Nonlinear probit model

[Maarten buis <maartenbuis@yahoo.co.uk>]

--- On Mon, 14/3/11, B Villena wrote:
> I have a non linear probit model that looks like this
> 
> y_i = F( ( a0 + a1*X1_i + a_2*X2_i)*( b_1*Z1_i +
> b_2*Z2_i))
> 
> where y_i is a binary dependent variable, and F(.) as
> normal CDF. 

That specification is almost certainly wrong. This is 
similar to an interaction model (the difference is that
you want to include a proportionality constraint). In
your specification you are leaving out the main effects
of X1 and X2. There are extremely rare circumstances 
where leaving out the main effects can be justified,
but they are so rare that the likelihood that this 
applies to your situation is de facto zero, unless you
have thought about this very carefully and delibaretly
left the main effects out.

Anyhow, I have sent an update of the -propcnsreg- package
to Kit that will estimate a correct version of the 
model you want to estimate. Actually it fits a logit 
instead of a probit, but I would consider that an 
improvement too ;-). As soon as Kit updated the package
I will announce it on the list.

Hope this helps,
Maarten

--------------------------
Maarten L. Buis
Institut fuer Soziologie
Universitaet Tuebingen
Wilhelmstrasse 36
72074 Tuebingen
Germany

http://www.maartenbuis.nl
--------------------------


      

*
*   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/