st: Biprobit , margin subsampling and margins

[Ayman Farahat <ayman.farahat@yahoo.com>]

Hello; 
I have a bivariate probit of the form 
(x = y z) ( y= a b)

x is a rare event ; that is there are way more negative than positive. 
y is also a rare event . 

The bivariate model will get a very high accuracy by setting all x to zero. 

I have been downsampling the negative x to get a balanced set. 
Question :
1) Does that make sense?
2) How to interpret the coefficients of the model ? for example the correlation between the errors "rho", how does one make sense of it after the subsampling.
3) How o i calculate ATE after subsampling?

Thanks
Ayman

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/