On Wednesday, September 24, 2003, at 02:33 AM, Robert wrote:
I have a large dataset that includes hundreds of thousands of
individuals in
over 1,200 different schools each observed for up to 6 time periods. I
would like to estimate a model with individual and school fixed
effects. Is
there anyway to control for school fixed effects without adding over
1,200
dummy variables (I don't think Stata will invert a matrix that large).
Why can't you just apply the fixed-effects transformation "by hand" for
the schools? That is, compute (y_{s,t} - y_sbar), where y_sbar is the
average value of y over that school's six observations, and likewise
for each x? Those data could then be used as 'raw' data for the
fixed-effects model at the individual level.
That said, this sounds like a multilevel model, and some of the
techniques specifically designed for multilevel modeling may be more
appropriate than a two-level fixed effects model.
Kit
*
* 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/
