st: Re: stesting across regressions

[baum <baum@bc.edu>]

--On Wednesday, August 21, 2002 2:33 -0400 Hans wrote:

> I wonder if there is a command to test if the coefficient on x1 in
> regression (1) is equal to the coefficient on the same x1 but in another
> regression (2), which is differently specified. I am applying a simple
> OLS using robust standard errors.
> Many thanks for your hinds.
> Hans
This question resurfaces on the list every so often. It is not a technical 
issue (does Stata do...) but an econometric issue: generally speaking, such 
a test cannot be done. If you can place both equations into a single 
estimation problem--for instance, by specifying a SUR (sureg) in which both 
equations appear--then it is well defined. But if you run two separate 
regressions, there is no way to estimate the covariance of the two 
estimated coefficients.

Alternatively, one could (in some cases) pool the data, and run a single 
regression in which both coefficients appear, with dummies dealing with the 
alternative specifications. This would also raise an issue of 
heteroskedasticity between the two forms of the model, but that could be 
dealt with. But one way or another, both coefficients have to be estimated 
in the same coefficient vector for such a test to be well defined.

Kit
*
*   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/