[R-sig-ME] extracting qAIC or quasi-likelihood from glmmPQL

PAUL, Mathilde m.paul at envt.fr
Tue Dec 4 09:05:11 CET 2012


I am an epidemiologist working with spatially correlated data. I am using glmmPQL to fit spatial generalized linear mixed models (and have no alternative options as this seems to be the only way to specify the spatial correlation as a random effect).
I read Ben Bolker's document online about quasi-AIC (http://cran.r-project.org/web/packages/bbmle/vignettes/quasi.pdf) but it seems that nothing has been implemented for glmmPQL.
Does anybody know how to extract quasi-likelihood from glmmPQL outputs ? (Or any way to get quasi-AIC from glmmPQL ?) I could not find any response in R-help archives.

Here is Ben Bolker's reply:
"The only thing I can think of is to hack the source of the glmmPQL function to remove the lines that explicitly erase the log-likelihood.
Even here I would want to be *very* careful and make sure that the AIC comparisons were roughly in line with those of other packages in cases where the functionality overlapped."

Many thanks,

Mathilde Paul
Toulouse Veterinary College (France)

More information about the R-sig-mixed-models mailing list