[R] for help about logistic regression model
Douglas Bates
bates at stat.wisc.edu
Tue Nov 21 20:17:31 CET 2006
I forgot that there is a much simpler explanation to your original
question of why the values of the likelihood, AIC and BIC for a model
fit by glmmPQL are shown as NA. This is a deliberate decision by the
authors of the MASS package. See the logLik method for the glmmPQL
class, available as
MASS:::logLik.glmmPQL
The object returned by glmmPQL is actually produced by the lme
function in the nlme package fitting a weighted linear mixed model to
the adjusted dependent variable. The parameter estimates and standard
errors for this linear mixed model are appropriate approximations to
those of the generalized linear mixed model but the likelihood isn't.
The authors of the MASS package replace this inappropriate value by NA
in the above-mentioned method.
More information about the R-help
mailing list