[R-sig-ME] lmer (binomial) and mcmcsamp.. and now?
dieter.menne at menne-biomed.de
Sat Dec 8 16:47:10 CET 2007
Dieter Menne wrote
> > I tried to run mcmcsamp on a problem of the form
> > The model is
> > lmer(RespB~Treat+Gender+(1|Pat),family=binomial)
> > Results look ok, in reasonable agreement with method="PQL", and in good
> > agreement with Göran's glmmML.
> > Running mcmcsamp() gave strange results. See
> > http://www.menne-biomed.de/uni/mcmc20000.pdf
> > http://www.menne-biomed.de/uni/mcmc2000.pdf
Douglas Bates <bates at ...> writes:
> This is a known issue with the current lmer and you should not trust
> the results.
I am in a reviewer-generated deep hole. He (or she) asked for contrasts computed
via glmm, which works well by combining glmmPQL and estimable (from gmodels).
However, (s)he mentioned that glmmPQL could not be valid (Agresti?), because
errors could be underestimated.
Is there currently any way to compute arbitrary contrasts with lmer/glmer other
than the default treatment contrasts in summary()? mcmcsamp cannot be used (see
above), parameter _contrast_ of lrm1 seems to be dead (note that it does not
complain about the following, even if you replace sloap by grump, and I have not
found an example using it):
lrm1 = lmer(RespB~Treat*Gender(1|Pat),
and neither gmodels' estimable nor multcomp do accept glmer results.
More information about the R-sig-mixed-models