 Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.

# Re: st: Question related to CDSIMEQ

 From Nick Cox To statalist@hsphsun2.harvard.edu Subject Re: st: Question related to CDSIMEQ Date Sat, 24 Sep 2011 01:40:19 +0100

```This post has now been posted three times, as Kwaku Damoah also posted it.

the Statalist policy: at most one repost.

Nick

On Sat, Sep 24, 2011 at 1:26 AM, Econ Stat <econstat@ymail.com> wrote:

> In my previous post, I inadvertently missed attaching my signature to the post. One of the statalist users drew my attention to that. I thank him and apologize to all the statalist users for this unintentional act. Please find my questions below and if possible kindly send me a reply...........
> I need your help on the following econometric issue:
>
> I am doing a simultaneous 2-step estimation (analyzed in Maddala 1983, Model 3, using CDSIMEQ program in STATA) as following:
>
> I have the following set of equations:
>
> Model 1:
> --------------
> y1= β1'x1 + δ1y2* +  ε1, ....................eqn(1)
> y2*= β2'x2 + δ2y1 +  ε2,  y2=1(y2*>0)....eqn(2)
>
> where y1 observed and y2 dichotomous.
> (1) I would like to know, can I use the estimates of y2* (i.e., y2*hat) from the 2nd stage regression of the above model (Model 1) and use it as an explanatory variable in a simple probit or ordered probit regression model?
>  Like for example in the following probit model,
> y3*= φ3'x3 +  γ3y2*hat +  η3,  y3=1(y3*>0)....eqn(3)
>  I have a valid theoretical justification for using the predicted value y2*hat (predicted from the 2nd stage of Model 1 i.e. simultaneous equations model) in the y3* equation but I am unsure about the empirical justification.
>
> (2) Also using the 2nd stage estimates of y2* (i.e., y2*hat from Model 1) as an explanatory variable, can I do a 2-step GMM estimation?
>