[R-sig-ME] AIC comparisons between lmer and glm

Ken Beath ken at kjbeath.com.au
Sat Aug 1 03:44:07 CEST 2009

Not necessarily. Some routines calculate incorrect log likelihoods and  
thus AIC by ignoring constants, but I'm not certain what the status of  
lmer is. It happens in other programs.

Maybe try some simulated data with a negligible random effect, then  
there should be no difference whether the random effect is included.


On 30/07/2009, at 2:11 AM, cacabelos at uvigo.es wrote:

> Hi R-users,
> I was unfortunately not able to find a solution to my problem about  
> model selection. I have fixed (e.g. Height) and random (e.g.Time)  
> factors, and I created these kind of model structures:
> Model1<-lmer(H~Height+(1|Time)+Biomass)+(Biomass| 
> Time),data=dat,family=gaussian)
> Model2<-glm(H~Height+Biomass,data=dat,family=gaussian)
> I am using the criterion ?the best model is the one that has the  
> lowest AIC?, but are these AIC from different procedures (i.e. glm  
> and lmer) comparable?
> I wonder if anyone can help me... Thank you!
> _______________________________________________
> R-sig-mixed-models at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models

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