Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.
From | "Mareike" <MareikeHahr@aol.com> |
To | <statalist@hsphsun2.harvard.edu> |
Subject | st: two-stage dummy variable model |
Date | Tue, 1 Jun 2010 19:19:42 +0200 |
Dear Statalist, I'm running a fixed effect model on panel data. I assume that one of my explanatory variables, a dummy variable taking the values 0 or 1, is endogenously determined. Thus I was planning to apply an instrumental variable technique in the form that I run a two-stage model: in the first stage I run a logit model, including as dependent variable the variable that was assumed to suffer from endogeneity in the original model and as regressors the exogenous variables from he main equation. In a second stage, the predicted probabilities from the logit should be interacted with the observed dichotomous values of the (endogenous) dummy variable and therefore used as the instrument in a two stage dummy-variable model. I got this idea from a paper from Heckman (1978) and think that it might solve my specific problem very well. But when trying to realize this idea, I really had problems of telling Stata what to do exactly. How can I tell Stata to use the predicted values from the logit model, interact them with the dichotomous values of the dummy variable and then to use these values for the endogenous variable in the fixed effect regression??? I would be very thankful if anybody of you can think of a convenient way of realizing this. Thanks a lot in advance Mareike * * 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/