Stata The Stata listserver
[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

Re: st: [Stata 9]: Bug in ivprobit, twostep?


From   "Brian P. Poi" <bpoi@stata.com>
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: [Stata 9]: Bug in ivprobit, twostep?
Date   Sat, 7 May 2005 20:23:14 -0500 (CDT)

On Sat, 7 May 2005, Arif Mamun wrote:

Hi,

In Stata 9, I am trying to implement ivprobit, twostep with the option "first". But the output doesn't provide any report of the first stage regression. Any idea about why this is happening?

On a related note, does the ivprobit, mle model take much longer to implement? In fact, I waited for couple of hours without luck for a specification that ran in a few minutes with the two-step model.

Like Scott Merryman reported earlier, I was unable to replicate Arif's
problem of not seeing the first-stage results when using the "first" and
"twostep" options.

However, if Arif emails me privately the exact syntax he was using and
a dataset to replicate the problem, I will be more than happy to look
into it.

Regarding the relative speed of the two-step and ML estimators, the
likelihood function is somewhat more complicated for probit with
endogenous variables than for the regular probit model, and so it
takes a bit longer to maximize.  This is especially true if the model
contains multiple endogenous variables.  Since the two-step estimator
is just linear regression followed by probit, it is typically faster
than the full MLE.


  --Brian Poi
    bpoi@stata.com


*
*   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/



© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index