[R-sig-ME] glmer.nb(): Residual term ??

Daniel Fulop dfulop.ucd at gmail.com
Tue Apr 28 17:21:44 CEST 2015


Hi Christine,

Others more knowledgeable than I may chime in, but as far as I know the 
GLMM parameterization that lme4 uses doesn't have a residual error 
term.  Other R software does have residuals, such as MCMCglmm.

HTH,
Dan.

Christine Adrion wrote:
> Hello,
>
> I fit a negative binomial GLMM using the lme4 package (version
> lme4_1.1-7), R-function glmer.nb(). It seems that this is still an
> experimental feature. The help function does not explain the residual term
> in the resulting R output.
>
> #---------------
> # Reproducible example:
> tmpf<- function() {
>      x<- runif(400) - 0.5
>      z<- gl(n=40, k=10)
>      m<- model.matrix(~x + z)
>      u<- rnorm(40, sd=0.5)
>      eta<- m %*% c(0, 3, u[-1] - u[1])
>      y<- rnbinom(n=length(eta), size=3, mu=exp(eta))
>      data.frame(y, x, z)
> }
>
> set.seed(2011)
> simdf<- tmpf()
> m<- glmer.nb(y ~ x + (1|z), data=simdf)
> m
>
> ##------------ output --------------------
> Generalized linear mixed model fit by maximum likelihood (Laplace
> Approximation) ['glmerMod']
>   Family: Negative Binomial(6.7839)  ( log )
> Formula: y ~ x + (1 | z)
>     Data: ..2
>        AIC       BIC    logLik  deviance  df.resid
>   890.5451  906.5109 -441.2725  882.5451       396
> Random effects:
>   Groups   Name        Std.Dev.
>   z        (Intercept) 0.3923
>   Residual             0.9665
> Number of obs: 400, groups:  z, 40
> Fixed Effects:
> (Intercept)            x
>      -0.5511       2.5969
> #-----------------------------------------
>
> Unfortunately, it's not clear to me what the measure "Residual" exactly
> expresses and thus why it is used by glmer.nb.
>
>
> [BTW, the fitted dispersion parameter 'size' is not very close to the true
> one (which was 3).]
>
> Thanks for any help/explanation.
>
> Kind regards
> Christine
>
> _______________________________________________
> R-sig-mixed-models at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models

-- 
Daniel Fulop, Ph.D.
Postdoctoral Scholar
Dept. Plant Biology, UC Davis
Maloof Lab, Rm. 2220
Life Sciences Addition, One Shields Ave.
Davis, CA 95616



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