[R] anova on binomial LMER objects

[Douglas Bates]

The issues with lmer and the analysis of variance are due to its not
make appropriate correction for the prior weights vector.  If you
convert your binomial response to the equivalent number of binary
responses you get an appropriate anova table.

It's on the "ToDo" list to fix this but a few other things have to
come first, like grading assignments in one of my courses and
repairing the computer in my office.  This is the third motherboard I
have torched in four months.



On 9/29/05, Alan Cobo-Lewis <alanc at umit.maine.edu> wrote:
> On Wed, 28 Sep 2005, Robert Bagchi wrote:
> >>Hi Patrick
> >>
> >>thanks for your advice. I have now tried glmmPQL, and it worked fine -
> >>I'm getting consistent results between plots and models fitted by
> >>glmmPQL. Plus it allows predict() and resid() which is another advantage
> >>over lmer at present.
> >>
> >>quick question though: why does one need to use PQL for binomial models?
> >>Is there a good reference for this?
> >
> You don't have to use PQL for binomial models, but you can't use least-squares. PQL is an approximate solution. Laplace and Adaptive Gaussian Quadrature options in lmer are better approximations. So lmer would likely become the better option as it
> progresses in its development (though the current issues you've found with the F ratios certainly sound like maybe lmer isn't better for you in its current incarnation).
> alan
>
> --
> Alan B. Cobo-Lewis, Ph.D.               (207) 581-3840 tel
> Department of Psychology                (207) 581-6128 fax
> University of Maine
> Orono, ME 04469-5742                    alanc at maine.edu
>
> http://www.umaine.edu/visualperception
>
> ______________________________________________
> R-help at stat.math.ethz.ch mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html
>