Skip to content

glmer != lmer(,..family=binomioal)

4 messages · Dieter Menne, Douglas Bates

Original
Dieter Menne
# 
Dear lmers,

from reading the docs and the first lines of lmer() code, I got the
understanding that the two formulations below should be equivalent. The first
works as expected, the second gives an envir() error. What's wrong?

Dieter


x=data.frame(resp=rbinom(100,1,0.8),
           treat=sample(letters[7:8],100,TRUE),
           pat=as.factor(sample(letters[1:4],100,TRUE))
           )
glmer(resp~treat+(1|pat),family=binomial,data=x)
lmer(resp~treat+(1|pat),family=binomial,data=x)
Douglas Bates
# 
Can you tell us the version of the lme4 package that you are using, please?
On Dec 3, 2007 7:54 AM, Dieter Menne <dieter.menne at menne-biomed.de> wrote:
Dieter Menne
# 
Douglas Bates <bates at ...> writes:
Bad, Dieter. Time to read the posting guide. 

This is my home-build version for Windows from r-forge. I admit that I forgot to
explicitly check for a fresh version of Matrix.

Dieter


Package: lme4
Version: 0.999375-0
Date: 2007-11-10
Title: Linear mixed-effects models using S4 classes
Author: Douglas Bates <bates at stat.wisc.edu>
Maintainer: Douglas Bates <bates at stat.wisc.edu>
Description: Fit linear and generalized linear mixed-effects models.
Depends: methods, R(>= 2.6.0), Matrix(>= 0.999375-1), lattice
LinkingTo: Matrix, stats
Imports: graphics, stats
Suggests: coda, mlmRev, MEMSS
LazyLoad: yes
LazyData: yes
License: GPL (>=2)
Packaged: Sun Dec 2 09:18:26 2007; theussl
Built: R 2.6.1; i386-pc-mingw32; 2007-12-03 11:36:46; windows
Douglas Bates
# 
Hmm.  Those calls should be equivalent in that version so I will look
into why they aren't.

At present I am involved in hand-to-hand combat with that code trying
to determine why the calculation of the Laplace approximation for the
glmer methods is not correct.  I thought I had it working but on
further checking I discovered that I don't.  Don't use that version
for glmer for a few days.
On Dec 3, 2007 8:10 AM, Dieter Menne <dieter.menne at menne-biomed.de> wrote: