Stata: Data Analysis and Statistical Software

Hi,

I think you can do stepwise regression-model-building as long as you bear in mind that, prior to having the complete model specification, you might have an omitted variable bias. So be careful when interpreting...

I would start with xtreg, buidling stepwise to the complete model with all factors assumed exogenous, and do xtivreg as last estimation. This also allows you to judge the bias from not taking into account endogeneity (could be small).

Justina


-----owner-statalist@hsphsun2.harvard.edu schrieb: -----

An: statalist@hsphsun2.harvard.edu
Von: Neil Shephard <nshephard@gmail.com>
Gesendet von: owner-statalist@hsphsun2.harvard.edu
Datum: 05.01.2011 12:04PM
Thema: Re: st: xtivreg- stepwise regression models

On Wed, Jan 5, 2011 at 10:42 AM, Ari Dothan <ari.dothan@gmail.com> wrote:
> Hi all,
> xtivreg requires the presence of an endogenous variable.
> Stepwise regressions, on the other hand, start with the controls in
> the absence of the IV.
> Since that endogenous variable is the regressor which should enter in
> the last hierarchical regression, and since this is not possible in
> xtivreg, I would much appreciate guidance concerning the correct way
> to run xtivreg stepwise regressions.

Not very helpful, but stepwise variable/model selection should be
avoided.  Some comments in this FAQ
http://www.stata.com/support/faqs/stat/stepwise.html

Neil

--
"Our civilization would be pitifully immature without the intellectual
revolution led by Darwin" - Motoo Kimura, The Neutral Theory of
Molecular Evolution

Email - nshephard@gmail.com
Website - http://kimura.no-ip.org/
Photos - http://www.flickr.com/photos/slackline/
*
*   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/