[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]
st: Re: xtrc -- Should the (mean) coefficients be different than those ofxtreg, fe, and if so, why and in what direction?
Wow - brief Statalist post ever? (just a subject)
1) Yes, xtrc coefficients can be expected to differ from xtreg, fe
2) Why? - because it's a different estimator? Read the manual (or any
3) Direction - it depends...
A little less briefly - the fe model is basically pooled OLS after de-meaning
within each panel. The xtrc model is more like a weighted average of the
coefficient estimates from individual regressions run within each panel
separately. Panels with the better-determined coefficient estimates get greater
weight in determining the overall (mean) coefficient estimate. So if the
coefficient for a certain variable tends to vary with the amount of "noise" in
the panel, then the difference can be systematic.
----- Original Message -----
From: "david reinstein" <firstname.lastname@example.org>
To: "statalist" <email@example.com>
Sent: Friday, June 01, 2007 12:22 PM
Subject: st: xtrc -- Should the (mean) coefficients be different than those of
xtreg, fe, and if so, why and in what direction?
* For searches and help try: