[R-sig-ME] [R] Question about GLMER

Douglas Bates bates at stat.wisc.edu
Tue Nov 16 23:48:52 CET 2010

On Tue, Nov 16, 2010 at 4:02 PM, Daniel Jeske <daniel.jeske at ucr.edu> wrote:
> Hi All -

> Doug, thanks for your reply.  The context I'm looking at is a Poisson
> GLMM with random (intercept,slope) for each subject.  The
> variance-covariance matrix is 2x2.  By unstructured, I meant a 3
> parameter matrix (sig1^2,sig2^2,sig12), as compared to a (reduced)
> alternative diagonal structure.

I thought that might be what you were doing. Look at the fm1 and fm2 from


fm1 allows for correlation of the random effects within subject.  fm2 doesn't.

> Dan
> -----Original Message-----
> From: dmbates at gmail.com [mailto:dmbates at gmail.com] On Behalf Of Douglas
> Bates
> Sent: Tuesday, November 16, 2010 1:01 PM
> To: Daniel Jeske
> Cc: r-help at r-project.org; R-mixed models mailing list
> Subject: Re: [R] Question about GLMER
> I am cc:ing the R-SIG-Mixed-Models at R-project.org mailing list on this
> reply as such questions are often answered more quickly on that list.
> On Tue, Nov 16, 2010 at 2:00 PM, Daniel Jeske <daniel.jeske at ucr.edu>
> wrote:
>> Dear R Help,
>> I believe the glmer() function in lme4 automatically fits an
>> unstrucruted covariance matirx for the random effects. Is that true?
> The short answer is "no".
> The longer answer is that you will need to be more specific about the
> form of the data and the model before your question can be answered.
> "unstructured" variance-covariance is SAS-speak (and an oxymoron, but I
> promise not to rant about that).
> So please give us a context.
>>    If so, do I have an option to somehow ask for a
>> diagonal structured covariance matrix?
> No virus found in this incoming message.
> Checked by AVG - www.avg.com
> Version: 9.0.869 / Virus Database: 271.1.1/3237 - Release Date: 11/15/10
> 23:34:00

More information about the R-sig-mixed-models mailing list