peixotop <peixotop <at> leuphana.de> writes:
> I am using glmmADMB and when I run some models, I recieve the following
> message:
>
> Erro em glmmadmb(eumencells ~ 1 + (1 | owners), data = pred3, family >
"nbinom", :
> The function maximizer failed (couldn't find STD file)
> Furthermore: Lost warning messages:
> Command execution 'C:\Windows\system32\cmd.exe /c
> "C:/Users/helenametal/Documents/R/win-library/2.15/
> glmmADMB/bin/windows32/glmmadmb.exe"
> -maxfn 500 -maxph 5 -noinit -shess' teve status 1
> : Mensagens de aviso perdidas:
> execu??o do comando 'C:\Windows\system32\cmd.exe /c
> "C:/Users/helenametal/Documents/R/win-library/2.15/
> glmmADMB/bin/windows32/glmmadmb.exe"
> -maxfn 500 -maxph 5 -noinit -shess' teve status 1
Sorry, this is not nearly enough information for diagnosis.
This message just means that *something* went wrong during the
optimization step (I do appreciate that it would be good to
improve the error messages, although there may not be that
much more information available).
Please (1) follow-up to r-sig-mixed-models at r-project.org
and (2) give more complete information on the full model you
ran, contents of pred3, etc. (see e.g. http://tinyurl.com/reproducible-000)
Here's a minimal example that shows that a model of the
form you present *could* work:
pred3 <- data.frame(owners=rep(letters[1:20],each=20))
set.seed(1001)
u <- rnorm(20,sd=2)
pred3$eumencells <- rnbinom(nrow(pred3),mu=exp(1.5+u),size=2)
library(glmmADMB)
glmmadmb(eumencells ~ 1 + (1|owners),family="nbinom",data=pred3)
-- although it doesn't work very well -- it essentially estimates
the random effects as zero, lumps the among-owner variance into
the NB variance, and mis-estimates the intercept. I don't blame
glmmADMB for this, though, it's a small data set and a tough problem.