[R-sig-ME] New alpha release of lme4 available on R-forge

Douglas Bates bates at stat.wisc.edu
Fri Feb 8 01:22:04 CET 2008

On Feb 7, 2008 5:53 PM, Austin Frank <austin.frank at gmail.com> wrote:
> On Wed, Jan 30 2008, Douglas Bates wrote:
> > - The effect of the link function and the variance function in the
> > family argument have been separated.  Families beyond the binomial and
> > poisson are now allowed although I would advise caution.  I need to do
> > some more research on whether a common scale parameter changes the
> > calculation of the deviance.  (Feature request #81).

> Dr. Bates--

> Reading through the feature request and your response to it, I was not
> quite clear on how to interpret this message.  Above you say "families
> beyond the binomial and poisson are now allowed...".  Does this mean
> that you've added the families and linking functions that were in the
> previous version of lme4 (gamma, etc.), or that any family may be used?

I should have been more explicit.  One can use any of the families
defined in the stats package for R, including the quasi family.
Essentially those families define the link and inverse link functions
and the variance function.  Because these functions can end up being
called many, many times in optimizing the log-likelihood for model
they are evaluated in compiled code.

They could be extended but that would need a definition of inverse
link, the derivative of the inverse link and the variance function for
any new families that are to be incorporated.

> Unsurprisingly, I'm wondering whether the above change will enable me to
> model a multinomial dependent variable using glmer.  Is this now
> possible?
> Thanks for your help,
> /au
> --
> Austin Frank
> http://aufrank.net
> GPG Public Key (D7398C2F): http://aufrank.net/personal.asc
> _______________________________________________
> R-sig-mixed-models at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models

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