[R-sig-ME] lme4 with Poisson
bbolker at gmail.com
Fri Aug 31 23:14:35 CEST 2012
Douglas Bates <bates at ...> writes:
> On Thu, Aug 30, 2012 at 5:25 PM, Lynne Clay <lynne.clay at ...> wrote:
> > Dear Prof Bates,
> > I'm a doctoral candidate in NZ trying to analyse survey data with random
> > effects with my outcome being a count. I discovered your lme4 package and
> > have been using this with success, however, I need to check for
> > overdispersion and it is at this point I am having problems. The formula I
> > have used before has been (1/df)*deviance and if I use this my model is
> > highly overdispersed. I read on one of the discussion boards that adding an
> > extra random effect (1|id#) addresses the overdispersion problem which I
> > have included but overdispersion continues.
> > Can overdispersion be calculated in this manner?
> I'm sorry but I know nothing about overdispersion. To me it is
> completely artificial because there is no probability distribution on
> which to base a statistical model with these properties.
> > Do you have any suggestions of how to deal with this?
> Sorry but I don't. I have taken the liberty of sending a copy of this
> reply to the R-SIG-Mixed-Models mailing list in the hope that readers
> of that list can help you.
It is probably worth checking out http://glmm.wikidot.com/faq ,
which has a variety of suggestions for handling overdispersion in lme4
(via adding observation-level random effect, as you suggest above) and
in other R packages: among other things, there are other packages such
glmmADMB that can fit negative binomial models with random effects.
One of the other responses mentions sabreR: from my brief
web-scrounging (e.g http://sabre.lancs.ac.uk/sabreR_coursebook5.pdf ),
it looks like sabre handles overdispersion by individual-level
random effects as well.
I might also recommend Zuur et al's book on mixed models.
More information about the R-sig-mixed-models