[R-sig-ME] updated version of examples from Bolker et al 2009 TREE paper

[Ben Bolker]

  I've just uploaded some (significantly) revised versions of the
supplementary online material from the Bolker et al 2009 _Trends in
Ecology and Evolution_ paper.

http://glmm.wikidot.com/local--files/examples/Banta_2011_part1.pdf
http://glmm.wikidot.com/local--files/examples/Banta_2011_part2.pdf
http://glmm.wikidot.com/local--files/examples/Banta_2011_part1.Rnw
http://glmm.wikidot.com/local--files/examples/Banta_2011_part2.Rnw
http://glmm.wikidot.com/local--files/examples/glmm_funs.R

  Major changes:

* part 1 (glmer fit) -- we now do the analysis with a lognormal-Poisson
model (i.e. observation-level random effects), rather than with a
quasi-Poisson family.  We conclude that our old results about the
variance in effects among groups were wrong (!! vindication for the
decision to drop "quasi" families from lme4 !!), and that in fact there
is generally insufficient information in the focal data set to infer
more than that there is variation among families (i.e. we can fit
multilevel models with variance among the intercepts at several levels,
but not with variance among the effects).
   The analysis is slightly streamlined and (we hope) clearer.

* part 2: we extend the example to show equivalent fits via
  * MCMCglmm
  * glmmADMB
  * glmmPQL
  * lmer, using log(1+x)-transformed data

In due course these PDFs will replace the old SOM on Elsevier's web
site. We would be happy to get feedback.


  Ben Bolker