[R-sig-eco] GLMM or GLM - fish capture success data

[Brian Inouye]

Hi Joana,

Just some cautions: either model could work, but you might need to go 
further with checking some model assumptions before deciding which 
structure to use.  If the proportions overdispersed then using the GLMM 
might be better.  Also, it is dangerous to compare AIC values that you 
get from different packages, and sometimes even from different functions 
within the same package!  Different packages treat the constants in 
likelihood functions in different ways, so that the AIC values will not 
be on the same scale.  When running mixed models, also check the REML 
option -- you shouldn't compare models fit with REML to those without 
(see Ben Bolker's glmm FAQ for explanations why).

-Brian