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

RE: st: Two models or one dummy


From   "David E Moore" <[email protected]>
To   [email protected]
Subject   RE: st: Two models or one dummy
Date   Mon, 13 Oct 2003 17:27:48 -0700

The short and direct answer to the original question is, "no."

The longer answer makes note of the fact that the two different approaches,
while estimating the exact same equations, do not yield identical standard
errors.  The dummy variable with interaction approach assumes equal error
variance across all observations while the separately estimated equations do
not.  The reason Stata provides no test for which specification is correct is
that you can't use the same data to test the equality of the variances of the
errors for each category of the dummy variable *and* estimate the equations
under the assumption that the null hypothesis (for the test) is true.
Hopefully, that made sense.

In any case, it seems to be common practice to use interaction terms to test for
cross-equation differences, though one should probably be explicit about which
assumption makes the most sense on a case-by-case basis and estimate the
standard errors (and tests) accordingly.


> -----Original Message-----
> From: [email protected]
> [mailto:[email protected]]On Behalf Of David
> Greenberg
> Sent: Monday, October 13, 2003 4:25 PM
> To: [email protected]
> Subject: Re: st: Two models or one dummy
>
>
> The significance of the interaction term should give results that are
> equivalent to a test for the significance of the difference of the
> coefficients when you estimate regressions on each subsample. The
> individual coefficients in such regressions may be significant, but
> that is not the issue; it is whether they are significantly different
> from one another. David Greenberg, Sociology Department, New York University
>
> ----- Original Message -----
> From: Shahbaz Ali Sheikh <[email protected]>
> Date: Monday, October 13, 2003 5:01 pm
> Subject: st: Two models or one dummy
>
> > Hi everyone,
> >
> > Is there any test in STATA that helps you decide, whether to run
> > two seperate
> > models by dividing the sample on the basis of an exogenous
> > variable or
> > include a dummy variable.
> >
> > I confused because when I run the regressions on a single model
> > including the
> > dummy, the interactions are insignificant. However, when I run
> > seperate
> > models by dividing the sample on the basis of the dummy taking 1
> > and 0, I get
> > completely different and signficant results.
> >
> > Is there a way in STATA to test when we should run seperate models
> > instead of
> > a including a dummy varible?
> > Thanks.
> > Shahbaz
> > _____________________________________________
> > PhD Candidate
> > Brandeis International Business School
> > Brandeis University, MS 032,
> >
> >
> >
> >
> > *
> > *   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/
> >
>
> *
> *   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/
>

*
*   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 1996–2024 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index