just to clarify, you're not really interested in the goodness of fit
between the two nested models. If you were, then logit would do better.
relogit drops fit to reduce bias, which we normally care more about.
on fit, i'd look at calibration plots like that in
http://gking.harvard.edu/files/abs/improv-abs.shtml and
http://gking.harvard.edu/files/abs/toe-resp-abs.shtml
Gary
On Wed, 23 Feb 2005, Jennifer Scott wrote:
Dear All,
I understand Relogit is not a likelihood technique, so can�t perform the
likelihood ratio test (LRT) like form the ordinary logit. But, are there
any other ways of doing smething similar to that from Relogit; I
simplely want to perform the goodness-of-fit test between two nested
models.
Thanks all.
J.S.
---------------------------------
Do you Yahoo!?
Yahoo! Mail - Helps protect you from nasty viruses.