[R-sig-ME] Likelihood test for random effect in a GLMM

Renwick, A. R. a.renwick at abdn.ac.uk
Thu Nov 27 14:51:20 CET 2008


Dear All
Many apologises for asking yet another question to the forum!
This time I am questioning the validity of using a likelihood test to test the significance of a random effect in a GLMM.  I am using GLMM for data with a binomial distribution and also with a poisson distribution.  I want to determine if the GLMM provides a better description of the data than the GLM therefore I used the following test:

#GLM
model<-glm(fleaburden~sex+width+sess+Nhat+alt+sex:width+sex:sess+sex:Nhat+sex:alt+width:sess+width:alt+sess:Nhat+sess:alt, family=poisson, data=flea, subset=-233)

#GLMM
bin<-lmer(fleaburden~sex+width+sess+Nhat+alt+sex:width+sex:sess+sex:Nhat+sex:alt+width:sess+width:alt+sess:Nhat+sess:alt+(1|LocTran), family=poisson, data=flea, REML=FALSE, subset=-233)

#use loglikelihood to test to see if sig difference between glm and glmm
as.numeric(2*(logLik(bin)-logLik(model)))
#940.5673
 pchisq(940.5673,1,lower=FALSE)
# 1.294078e-10

However, the p value obtained seems to exceedingly low and thus I was wondering if this is a reliable test to preform.

Many thanks,
Anna

Anna Renwick
Institute of Biological & Environment Sciences
University of Aberdeen
Zoology Building
Tillydrone Avenue
Aberdeen
AB24 2TZ


The University of Aberdeen is a charity registered in Scotland, No SC013683.




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