> I'm using xtprobit, pa with a robust option in order to estimate a panel
> data model in which my dependent variable is the decision of firm to
> invest(1) or not (0) in 10 different countries (in 7 different years)
>
> I'd like to try different correlation structures, other than
> exchangeable, like, for example, an unstructured one to see what
> happens.
>
> When i write corr(uns) stata says that my time variable has duplicate
> values in the panel. This is true, actually, because i have 944 firms
> (my id) investing in 10 different countries form 1992 to 1998. So, my
> clusters are repeated per each country. But why stata doesn't give me
> this error when i run xtprobit, pa i(id) ?
With -corr(exch)-, it does not have to bother to identify to which
year an observation belongs. With -corr(uns)-, it does need this
information to link the observations together. -corr(uns)- is hard to get
converged though.
> How can i set my panel then? Tsset doesn't work.
I would suggest migrating to -gllamm- and have two random effects: one for
the firm, and one for the country. Otherwise, your model does not make
much sense: there is probably as much of the country effect as there is of
the firm effect, and you might as well want to run a model where the
country is your panel variable. This would be a two-way panel data set up.
(If you go as far as -gllamm-, you might as well drop in a third random
effect for the year, just in case.)
> Also, how is it possible that even with the option robust the
> correlation is always the same, i.e. the default one (exchangeable)?
I am not quite sure as to how exactly the -robust- option is to be
interpreted here. I suspect it does not do much, as any correlation
implies a -cluster-ed covariance matrix estimator that is overriding
-robust-. But that deserves more thinking.
--- Stas Kolenikov
-- Ph.D. student in Statistics at UNC-Chapel Hill
- http://www.komkon.org/~tacik/ -- [email protected]
* This e-mail and all attachments to it are not intended to provide any
* reasonable point of view and was transmitted to you in error. It
* should be immediately deleted by all recipients unless they really
* enjoy communicating with the author :). Other restrictions apply.
*
* 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/
