Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.
From | "Gregory, Christian" <cgregory@ers.usda.gov> |
To | <statalist@hsphsun2.harvard.edu> |
Subject | st: Binary Endogenous Interactions |
Date | Tue, 8 Mar 2011 16:50:27 -0500 |
Colleagues, I have a two equation model: Y1=zb + xb + e Y2 = xb + y1b + y1*v*b + u b is generic for coefficients. Both y1 and y2 are binary. Z are instruments for y1. I'm interested in the marginal effect of the interaction y1*v on Y2. I have estimated the above using ml and a method that deals with the correlation of e & u. I haven't instrumented for the interaction of y1 and v. The distribution of y1*v is bimodal-obviously, with a lot of zeros; conditional on y1==1, it is normal. So, a third equation-that predicts y1*v as a function of z-is possible, although I'm not sure the form it should take, and certainly not what distribution is appropriate to specify in the likelihood function. Any suggestions would be greatly appreciated. Thanks for your consideration. Christian * * 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/