The choice of generalized linear mixed models is difficult, because it involves the selection of both fixed and random effects. Classical criteria like Akaike’s information criterion (AIC) are often not suitable for the latter task, and others which are useful in linear mixed models are...