[R-sig-ME] observation level random effects; estimated variance of variance component estimates

[Jarrod Hadfield]

Hi,

I'm not sure how to fit the model you want in lmer, but hopefully  
someone will have a better idea.  However, if its of interest or use  
I've released the package MCMCglmm that fits the model you're after.

  By default MCMCglmm fits rand as a residual component:

m1<-MCMCglmm(Seizure ~ Base + Trt*Base + Age + V4, random=~id,  
family="poisson", data=dat)

which I believe is the model you want.

HPDinterval(m1$VCV)

gives CI's on the variance components (the second of which is you rand  
term)

Hope this helps,

Jarrod


-- 
The University of Edinburgh is a charitable body, registered in
Scotland, with registration number SC005336.