[R-sig-ME] Are there sampling methods for glmer objects?

[Ben Bolker]

Joshua Wiley <jwiley.psych at ...> writes:

> 
> Hi All,
> 
> The subject pretty much says it all.  Using: lme4_0.999999-0
> 
> >From the archives, I have seen Ben Bolker mention a function "glmersim".
> 
> ?mcmcsamp seems to indicate it can take model objects from glmer, but
> when I try it with a simple random intercept logit, I get:
> Error in validObject(.Object) :
>   invalid class "merMCMC" object: Slot sigma must have length 1.
> 
> I also seem to recall something about a "bootMer" function to do
> (faster) bootstrapping.

  See ?"simulate-mer", which replaces glmersim.
  I've just added a note on the mcmcsamp status to the FAQ:

http://glmm.wikidot.com/faq#mcmcsamp_status

(That said, the error message you got does seem a bit odd -- but
I'm not motivated to pursue it since I don't think it will work
properly anyway).

  Note that arm::sim is also available (although I'm not sure
what algorithm it uses, and it only samples beta, not theta [variance
component] parameters).