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]

Re: st: Quantile Regression with FE


From   Travis Smith <travisa.smith7@gmail.com>
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: Quantile Regression with FE
Date   Tue, 19 Feb 2013 07:26:05 -0600

G.A.Hughes@ed.ac.uk wrote:
> By removing the fixed effects you would be estimating conditional median deviations from the mean/median values of the observed variables.

To clarify Gordon's point, which is the crux of the issue, when you
demean, difference, or include an additive individual ``fixed'' effect
in a quantile regression, the estimated coefficients will NOT have the
same interpretation as the standard cross-sectional quantile
regression.

This is not to say this approach is wrong, but many people interpret
the estimation incorrectly.

See David Powell's recent work and references therein:
http://www.rand.org/pubs/working_papers/WR710-2.html

qswanqui@utk.edu wrote:
> Is there a more efficient way to estimate this similar to -xtreg-?

In the above article, you will see that Powell uses GMM. Under the
Powell specification, ``fixed effects'' are never specified, which
substantially reduces the number of parameters needed to identify the
model (in his example, from 162 to 1).  There is no canned package for
this approach and will require knowledge of GMM estimation using
Stata's -gmm- command and possibly with use of the moment-evaluator
program.  See Stata's reference manual for detailed examples.

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


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