
alternative param of NB
more complex random effects structure
  * how flexible is/should be possible?
  * interface/specification language: MCMCglmm, lme4, SAS PROC MIXED?

"R-side" random structure: temporal, spatial autocorrelation
  (hierarchical)

alternative distributions: binomial (n>1)

alternative links

alternative random effect distributions?

predict/simulate methods (with/without random effect specifications)
confidence intervals on predictions?

allow offset (group?) to be specified without quotes
(parse(substitute())?)