[R] glmmPQL

ripley@stats.ox.ac.uk ripley at stats.ox.ac.uk
Mon Jul 1 14:24:06 CEST 2002

On Mon, 1 Jul 2002, Christof Bigler wrote:

> Dear R users,
> can anybody explain me why the function glmmPQL(.) behaves in different
> ways, depending on the number of measurements/individuals you use? To
> show you this, I generated two examples. The first one includes 20
> indivduals with each 100 repeated measurements (binary response), the
> second one includes 40 individuals. The 'individuals' differ only in
> different x values. I fitted logistic regression models with and without
> random intercepts.
> The coefficients and p values between dummy.glm20 and dummy.glmm20
> differ. However, dummy.glm40 and dummy.glmm40 have the same coefficients
> and p values, respectively. Did the solution in the second example not
> converge (only one iteration step)?
> Why does the AIC between e.g. dummy.glm20 and dummy.glmm20 differ so
> much?
> And last question: how can dummy.glm20 and dummy.glmm20 be compared with
> an anova(.) function?

You can't use either AIC or anova with glmmPQL: it is not maximum
likelihood but PQL, which is an appoximate method.

You should remember that glmmPQL is part of the MASS package (which you
maybe didn't realise), to support a book you can't have read as it is not
yet published.

Brian D. Ripley,                  ripley at stats.ox.ac.uk
Professor of Applied Statistics,  http://www.stats.ox.ac.uk/~ripley/
University of Oxford,             Tel:  +44 1865 272861 (self)
1 South Parks Road,                     +44 1865 272860 (secr)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595

r-help mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To: r-help-request at stat.math.ethz.ch

More information about the R-help mailing list