[R-sig-ME] robust linear mixed model

Ben Bolker bbolker at gmail.com
Tue Oct 9 14:16:47 CEST 2012

Geraci, Marco <m.geraci at ...> writes:

>  You might want to consider a median regression with random
> effects. See the lqmm package
> http://cran.r-project.org/web/packages/lqmm/index.html. You can
> extend the analysis to other quantiles as well.
  This sounds like a good idea (better than anything I could come
up with).  I was going to suggest using WinBUGS or AD Model Builder
to construct a mixed model with t- rather than Gaussian-distributed
residuals (this could eventually be built into glmmADMB, but at
the moment you'd have to learn ADMB and build your own model
from scratch).

> -----Original Message-----
>  I have a dataset with some outliers or influential observations,
> and intend to do a mixed model regression.  I am looking for
> procedures or a package that can do "robust linear mixed model", to
> account for the outliers or influential observations or groups.
> There are similar functions for fixed models (rlm) in MASS package,
> but I haven't been able to find anything on mixed models.

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