My understanding of this is that it depends on the number of
observation you have in each group. How much is enough? I'd really
like to hear what other people think about this or if there are
specific references on the issue.
Anyway, see below for a post on similar issues by Gould, Wiggins and
Drukker of Stata Corp. where, if I well understood, the suggested
number of observation within each group should be greater than 50. As
a rule of thumb, they suggest also that the use of dummies
is allowed
"when you have enough observations in each group that you would be
willing to use the estimator on each group BY ITSELF and publish those
results in isolation".
See here:
http://www.stata.com/statalist/archive/2003-09/msg00103.html
How to explain this to a reviewer is another story...
Mirko
On 8/27/07, CASTELLUCCI Fabrizio <[email protected]> wrote:
> I'm running an ologit on an unbalanced panel. the DV has 7 categories,
> T=22, and I have 3182 observation. Among other estimators, i included
> 21 dummies as fixed effects. One reviewer suggested that, by doing so,
> i have an incidental parameter problem. I though that given my panel
> size the problem should not be that serious, but i was wondering
> whether there are ways to tackle such problem.
>
> thanks
>
> Fabrizio
> *
> * 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/
