I had a similar problem a few months ago using Hausman with xtlogit. As I recall, xtlogit was storing the e(b) and e(V) matrices using names like y: x1, y: x2, whereas logit was storing them as simply x1 , x2. The coefficient matrices (and e(V)) matrices had different column names, and appeared to have different variables. To get around this, I extracted the relevant submatrices and constructed my own Hausman statistic, though perhaps someone knows a more elegant way of doing this - perhaps by renaming the columns in the matrices.
Kanaka
RE:
Subject: st: hausman test for fixed effect logit vs. simple logit
I am trying to test whether a fixed effect logit (xtlogit) is a better
fit than simple logit (logit). An error appeared when I did the
hausman test. Could someone help me out? Thank you!
xtlogit y x1 x2 x3, fe i(igrp)
est store panel
logit y x1 x2 x3
hausman panel
. no coefficients in common; specify equations(matchlist)
for problems with different equation names.
r(498);
Yun
<<winmail.dat>>