|
[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]
st: reference for simulation of biased estimates with logit and dummies
From |
"Rosy Reynolds" <[email protected]> |
To |
"statalist" <[email protected]> |
Subject |
st: reference for simulation of biased estimates with logit and dummies |
Date |
Thu, 7 Feb 2008 09:22:51 -0000 |
Hi,
I have data where I expect clustering by centre, and have read repeatedly on
Statalist that adding centre dummies to represent fixed effects produces
biased estimates unless the number of observations in each centre is large.
David Harless in Oct 2007
(http://www.stata.com/statalist/archive/2007-10/msg00926.html) helpfully
pasted a detailed explanation from Bill Gould and Vince Wiggins from Feb
2000, which includes mention of a simulation study showing that the bias
really matters: "In the case of logistic regression, however, the estimates
one obtains from including all the dummies are biased and, even as
n->infinity, that bias never goes away. Vince Wiggins <[email protected]>
and I recently simulated this and discovered that this not a sterile,
theoretical argument -- the estimates on obtains for the parameters are
genuinely bad." I think the same simulation was alluded to again in 2003
(http://www.stata.com/statalist/archive/2003-09/msg00103.html) and 2007
(http://www.hsph.harvard.edu/cgi-bin/lwgate/STATALIST/archives/statalist.0710/date/article-934.html).
I would like to cite this simulation, or something similar, but have not
managed to find any more detail. Could anyone please direct me to a
reference or confirm that it was internal work and not formally published?
(I was also unable to find the Feb 2000 Statalist entry. It seems that the
archive does not go back so far. Is that right?)
Thanks to anyone and everyone who can help.
Rosy Reynolds
*
* 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/