Re: st: Newey's Two-Step Estimator & ivprobit postestimation

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

On Fri, 16 Jun 2006, Talip Kilic wrote:

> Hi,
>
> I am estimating an ivprobit model, in which I am instrumenting for two
> endogenous regressors. As suspected and stated in the STATA manual, MLE,
> the default option, does not lead to convergence and iterations never
> end if one wants to instrument for more than one endogenous regressor in
> the ivprobit model. Once the Newey's two-step estimator option is
> specified, the ivprobit model with multiple endogenous regressors is fit
> unproblematically. However, ivprobit postestimation tools are also not
> available under the two-step option. In this respect, I am wondering
> whether any one has had any experience with postestimation under
> ivprobit two-step option and if so, would you be able to provide any
> hints, suggestions or resources that would lead me to estimate marginal
> effects in my case? Any guidance would be appreciated. Thank you.
>
> Best,
> Talip
Talip,

The fundamental issue is that the two-step estimator only estimates the 
parameters of the probit equation up to a scalar multiple of their 
population values.  We can still test whether the two-step estimate of the 
coefficient on the endogenous regresssor is significantly different from 
zero, but we cannot use the estimates for postestimation analyses such as 
calculating marginal effects with the -mfx- command.

All is not lost, however.  Wooldridge (2002, p. 475) does mention two ways 
that average partial (ie. marginal) effects can be obtained from the 
two-step estimator.

Hope this helps

   -- Brian Poi
   -- bpoi@stata.com

-----

Wooldridge, J. M. (2002). Econometric Analysis of Cross Section and
    Panel Data.  Cambridge, MA: MIT Press.
*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/