Stefano Costalli <[email protected]>:
I think the approach is ill-advised, but you can
ssc inst fese
help fese
for a program to save estimated fixed effects and SEs from a FE
regression, then specify the SE as aweights in your second regression
(weighting the second step by the precision of the estimates from the
first step). The standard errors in the second regression will still
be wrong, so a Murphy-Topel correction [see SJ 6(4):521--529] or
-bootstrap- may be required.
Is this designed to recover a RE-type estimate? Presumably when
-hausman- rejects RE? You may want to try -xthtaylor- here instead.
On Fri, May 2, 2008 at 12:43 PM, Stefano Costalli
<[email protected]> wrote:
> Dear all,
>
> I'm searching for ways to obtain coefficients of time invariant variables in a fixed effects model. I have tried with -xtfevd-, a command recently written and discussed on this forum, but I'd like to do some comparisons.
> I found a paper where the author says that he calculated the coefficients of time invariant variables in two steps. First, he runs the fixed effects model with all indep. vars. and then he runs an ols regression of the time invariant variables on the results of the fixed effects (that in this step embody the dep. var.). I have some problems in understanding this step.
> Could someone help me? How would you transtale it in stata language? I mean, the first step is easy (-xtreg- with -fe- option), but then how should I save the results of the fixed effects? And, finally: do you think the solution he proposes is correct?
*
* 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/
