[R-sig-ME] correlation structure with glmer?

[Alice Shelly]

Very good point, and it makes me wonder why and how other methods (like 
glmmPQL) are
providing results.
Thanks so much for your response and all your work in R!
Alice
----- Original Message ----- 
From: "Douglas Bates" <bates at stat.wisc.edu>
To: "Alice Shelly" <alice at blarg.net>
Cc: <r-sig-mixed-models at r-project.org>
Sent: Monday, November 23, 2009 2:08 PM
Subject: Re: [R-sig-ME] correlation structure with glmer?


> On Mon, Nov 23, 2009 at 1:32 PM, Alice Shelly <alice at blarg.net> wrote:
>> Hello-
>> Is there any way to introduce correlation structure into a glmer model?
>> I tried using
>> correlation = corAR1(form= ~age|plot)
>> as you would in lme, but I was ignored (so rude).
>
> Well, first you would need to decide how to define such a model.  It
> is not trivial.  For a linear mixed-effects model you can separately
> modify the specification of the mean and the variance of the
> conditional distribution of the response given the random effects.
> For common generalized linear mixed models (family = binomial or
> family = poisson) you can't.  Once you have the conditional
> independence assumption and you specify the conditional mean you have
> completely specified the conditional distribution.
>
>