[R-sig-ME] comparing models with quasipoisson family

Seyed Reza Jafarzadeh srjafarzadeh at gmail.com
Fri Mar 23 00:26:58 CET 2007


I am using 'lmer' to fit generalized linear mixed-effects models. I
was just wondering how I can compare two models with identical
fixed-effect terms but different random effects specifications  when I
am using the 'quasipoisson' family.

Can I still use 'anova' to compare them? Should a model with
'quasipoisson' family specification produce a logLik value or AIC?

How about the estimate of dispersion parameter as we can get by 'glm'?


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