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 at the end of May, and its replacement, statalist.org is already up and running.


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

st: dealing with high dimensional fixed effects


From   statalist <statalistrw@gmail.com>
To   statalist@hsphsun2.harvard.edu
Subject   st: dealing with high dimensional fixed effects
Date   Sat, 21 Aug 2010 14:59:35 -0400

I am looking for suggestions for estimating LSDV/fixed effect models for which there are potentially large numbers of fixed effects. 

I have a very large dataset, and estimating all these parameters, even after differencing out one set (areg/xtreg), is not possible on my UNIX system. All this being said, I am not really interested in the parameter estimates as much as controlling for unobserved heterogeneity. 

Do people have suggestions for this? I know about the various two-way FE estimators (felsdvreg, a2reg, gpreg), but these are missing various features that I am interested in preserving (namely clustering and weighting). One could possibly cluster ("wild bootstrap") with these estimators, but weighting is not possible as far as I can tell. 

Perhaps there is any easier way? I have not been able to reproduce estimates using a two-step, Frisch Waugh Lovell approach with weights, and I am wondering if I am missing something here...

xtreg y [aweight=weight], i(workerid) fe
predict y_res, ue
xtreg x [aweight=weight], i(workerid) fe
predict x_res, ue
xtreg y_res x_res [aweight=weight], i(firmid) fe cluster(group)

Any other suggestions for dealing with these types of (seemingly more and more common) issues? 


*
*   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