[R-sig-ME] MCMCglmm for binomial models?

David Duffy davidD at qimr.edu.au
Mon Sep 20 06:51:52 CEST 2010


On Sun, 19 Sep 2010, Bob Farmer wrote:

>  nitt = 50E3
>
> (with extra iterations for insurance), the results are far more
> variable, and generally not consistent with glmer()'s output.

You need to look at the effective sample size for each parameter (eff.samp 
in the summary output) -- 50000 reduces to only 10-80 independent 
observations for this example as I ran it, so there is still a largish 
Monte Carlo error.

> Assuming that glmer()'s output is okay (and maybe this is a big
> assumption?

Not for this simple problem.

Cheers, David Duffy.




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