Bookmark and Share

Notice: On March 31, it was announced that Statalist is moving from an email list to a forum. The old list will shut down on April 23, and its replacement, statalist.org is already up and running.


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: st: xtlogit, fixed effects, computation time


From   Xiang Ao <xao@hbs.edu>
To   "statalist@hsphsun2.harvard.edu" <statalist@hsphsun2.harvard.edu>
Subject   Re: st: xtlogit, fixed effects, computation time
Date   Wed, 17 Mar 2010 10:25:41 -0400

Your original specification

xtlogit export y_r lMA_ind yeard2-yeard11, fe

already has fixed effect on id, which is the combination of country, sector, and province. In my opinion, you should not put in province and sector dummies in, since id is a finer level.

Xiang

joachim jarreau wrote:
Hi statalist,

Im am trying to estimate a logit regression on a panel dataset with
four dimensions: province i (origin), country j (destination), sector
k and time (year) t. The estimated equation looks like:
P(Xijkt =1) = f (Ait, Bikt, Dijk, Dt)

where Ait are controls varying by province and time, B controls at the
province, sector, time levels, Dijk (ijk) are province-sector-country
dummies and Dt are time dummies.   In stata I do :

egen id=group(country sector province)
xtset id year
xtlogit export y_r lMA_ind yeard2-yeard11, fe

and it works fine - takes about 1/2 hours.
Now I would like to use different fixed effects: not a pure "within"
estimator like above, but one with separated fixed effects for the
province, sector and country dimensions. To do this I defined
"country" as the panel variable in xtset, and added province and
sector-dummies by hand, like:

xtlogit export y_r lMA_ind yeard2-yeard11 provd* sectord*, fe

and my problem is that this takes much too long: after about 20 hours
there was not one iteration displayed yet . I suppose the reason is
that when adding dummies in the regression line, these variables are
added to the list of regressors, whereas when they are associated to
the panel variable, stata only has to de-mean the variables within
each group, so that the regressor matrix size is not increased ?  What
can I do about it, is there a way to have the specification with
separated fixed effects estimated in reasonable time?


thank you in advance for your help.


joachim
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/


© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   Site index