[R-sig-ME] Do glmer and glmmadmb calculate log likelihood on thesame scale ?

[Simon Chamaillé-Jammes]

Thanks David and Joshua, 
my model was for binomial family. I've just checked on a simpler model
and logLik is exactly similar between glmer and glmmadmb. So if one use
the same model design (accounting for things like those Joshua revealed)
one should be able to compare the logLik and deviance between the two
implementations.

simon

On Thu, 2012-06-21 at 17:24 -0700, Joshua Wiley wrote:
> Just a note that those are not the same models.  To match glmer() with
> glmmadmb(), try:
> 
> gm2 <- glmer(y~Base*trt+Age+Visit+(1 | Visit) + (0 + Visit|subject),
>   data=epil2, family=poisson())
> logLik(gm2)
> 
> which for me gives:
> 
> 'log Lik.' -421.1128 (df=8)
> 
> OR you could change glmmadmb() to
> 
> fm2 <- glmmadmb(y~Base*trt+Age+Visit+(Visit|subject),
>   data=epil2, family="poisson", corStruct = "full")
> 
> however, on my machine, there are warnings about the matrix not being
> positive definite, and the model never converges.  If it worked for
> you, I would be curious what version of glmmADMB you are using.  Here
> is my sessionInfo()
> 
> R Under development (unstable) (2012-05-22 r59410)
> Platform: x86_64-w64-mingw32/x64 (64-bit)
> attached base packages:
> [1] stats     graphics  grDevices utils     datasets  methods   base
> 
> other attached packages:
> [1] lme4_0.999375-42  Matrix_1.0-7      lattice_0.20-6    glmmADMB_0.7.2.12
> [5] R2admb_0.7.5.2    MASS_7.3-18
> 
> loaded via a namespace (and not attached):
> [1] compiler_2.16.0 grid_2.16.0     nlme_3.1-104    stats4_2.16.0
> [5] tools_2.16.0
> 
> Cheers,
> 
> Josh
> 
> On Thu, Jun 21, 2012 at 2:22 PM, David Duffy <David.Duffy at qimr.edu.au> wrote:
> > On Thu, 21 Jun 2012, Simon Chamaillé-Jammes wrote:
> >
> >> I would like to know if glmer and glmmadmb calculate log likelihood /
> >> deviance on the same scale.
> >>
> > What model family?
> >
> > data(epil2)
> > epil2$subject <- factor(epil2$subject)
> > fm2 <- glmmadmb(y~Base*trt+Age+Visit+(Visit|subject),
> >                data=epil2, family="poisson")
> > gm2 <- glmer(y~Base*trt+Age+Visit+(Visit|subject)
> >             data=epil2, family=poisson())
> > logLik(fm2)
> > 'log Lik.' -655.41 (df=8)
> > logLik(gm2)
> > 'log Lik.' -272.4573 (df=9)
> >
> >
> > _______________________________________________
> > R-sig-mixed-models at r-project.org mailing list
> > https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models
> 
> 
>