Statalisters--
Thanks to Kit Baum, a new package -fese- to compute the standard
errors (SE) of fixed effects (FE) is available from SSC. To install
or reinstall -fese-, type
ssc install fese, replace
The data must be -xtset- or equivalent. The results are saved as
variables that vary only across panels, not within, and are equivalent
to ordinary linear regressions run on an exhaustive set of indicator
(dummy) variables for panels (with the -hascons- option).
Ordinary (OIM), heteroskedasticity-robust, and cluster-robust SEs are
calculated, but the usual asymptotic justification of cluster-robust
SEs does not apply, so the last should not be used except for research
purposes.
See the help file for more detail.
For large problems (many FE, severely unbalanced panels), the
calculations may take a long time, but you can request that only OIM
SEs be calculated with the -oonly- option.
You can set up a bigger problem (4710 FE) thus:
webuse nlswork, clear
xtset id
and on a 2.66GHz dual-core machine running Stata9/IC,
fese ln age, s(all_)
takes about 10 minutes, and
fese ln age, s(oi_) oonly
takes about 1 minute.
The exercise was suggested by Sarah Cohodes and others:
http://www.stata.com/statalist/archive/2008-01/msg00918.html
http://www.stata.com/statalist/archive/2007-01/msg01046.html
http://www.stata.com/statalist/archive/2006-04/msg00026.html
and the solution was suggested by Jesse Rothstein (not on Statalist).
*
* 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/
