--- In statalist@y..., "Ioan Voicu" <VOICUI@j...> wrote:
> I am trying to construct a within estimator for unbalanced panel.
> I have a two-way fixed effect. Both dimensions of the fixed effect are very
large (approx. 50,000 each), so I can't use dummies for any of the two.
>
> Is there a STATA procedure that does this?
> Alternatively, how can one transform the variables in order to estimate a
> within estimator using, for example, reg procedure?
Peter Davis has a paper that discusses multiway unbalanced panel methods.
He discusses both fixed and random effects methods. His methods may
require some clever linear algebra for problems as
large as yours.
Estimating Multi-Way Error Components Models with Unbalanced Data
Structures, Journal of Econometrics, 2002, Vol. 106(1), pp67-95. To be
reprinted in "Recent Developments in the Econometrics of Panel Data" edited
by Badi H. Baltagi and George Sumney, Edward Elgar Publishing, UK.
Another, perhaps more practical option is to use a conjugate gradient
method adapted into economics by Abowd and Kramarz. It originally was
developed for animal genetics, I believe.
http://instruct1.cit.cornell.edu/~jma7/abowdcv.html
Stata is not a robust enough problem to handle either of these
methods, unfortunately.
Jeremy
--
------------------------
Jeremy T. Fox
[email protected]
*
* 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/
