Re: st: about ivprobit

["Brian P. Poi" <bpoi@stata.com>]

On Thu, 31 Jul 2008, wangxin wrote:

> After using the ivprobit, how can stata report the
> predicted value for the first stage? Specifically, the
> predicted value for the endogenous variable.
Xin did not mention whether the maximum likelihood or two-step version of 
-ivprobit- is being used, so I'll address both.

For the maximum likehood (default) version, although -predict- after 
-ivprobit- does not have an option for the predicted values for the 
reduced-form equation for the endogenous regressor, they can be obtained 
using -matrix score-:

. sysuse auto
. ivprobit foreign mpg (gear = turn head)
. matrix b = e(b)
. matrix score ghat = b, eq(#2)

In the coefficient vector e(b), the first equation is the main response 
equation, and the second equation contains the equation for the endogenous 
regressor.  If you have, say, two endogenous regressors, then the 
coefficients are stored in equations two and three.

For the two-step estimator, e(b) only contains the coefficients for the 
main equation.  However, in this case the "first-stage" equations are fit 
using OLS, so you can just re-fit those equations using -regress-, then 
use -predict- afterwards.

   -- Brian Poi
   -- bpoi@stata.com
*
*   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/