[R] glmmPQL in 2.3.1

[Ben Bolker]

Christian Gold <c.gold <at> magnet.at> writes:

> 
> I have come across the previous communication on this list in September 
> (copied below) because I had received the same error message.
> 
> I understand from Brian Ripley's reply that anova should not be used 
> with glmmPQL because it is not an adequate method, and that this is now 
> shown with an error message.
> My question is, what method *should* be used? Using summary does not 
> give me the result I want, because it shows the significance of each 
> combination of factor *levels*, rather than factors, which can quickly 
> become very hard to interpret.
> 

 use lmer (in the lme4 package) with the "family" argument
set, to do real (or restricted) rather than quasi-likelihood
estimation?

   Ben Bolker