Bookmark and Share

Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

st: xtmixed, weights not allowed; xtmelogit, weights allowed?

From   Lucas <>
Subject   st: xtmixed, weights not allowed; xtmelogit, weights allowed?
Date   Wed, 6 Jun 2012 07:18:29 -0700

I estimate a model in xtmelogit.  I estimate the same model using
xtmixed, or, at least, I try. But, while I obtain estimates in
xtmelogit, for xtmixed I get "weights not allowed."  Yet, the manual
says fweights (and other kinds of weights) are allowed.  What's going

I provide this background for the curious. This is not part of my
question.  I am estimating the same model on 6 different groups.  I
have tried to do this simultaneously, but that xt model seems as if it
will never converge.  So, my fallback is to estimate on each group
separately -- note, I still have a grouping structure within the 6

I successfully obtain convergence in 10-20 iterations for 4 of the
groups. Two of the groups the Hessian becomes unstable.  I decide
perhaps it is a problem that different starting values might avoid.  I
think, perhaps I can obtain (scalings on) starting values by using
xtmixed -- this is akin to an old approach used by some older programs
when computing power was lower.  However, just to check things are
working, I attempt to estimate xtmixed even on a group for which I
have xtmelogit estimates.  The xtmixed model refuses to start,
reporting back  that "weights not allowed."

PS--If anyone has other ideas for how to fix the Hessian becomes
stable problem, I appreciate that.  However, my main query for now is
why xtmixed won't use weights when the same model runs fine using

*   For searches and help try:

© Copyright 1996–2017 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   Site index