[R-sig-ME] Repost in plain text: Discrepancy in residuals in equivalent glmer (lme4) and glmmadmb models

Ben Bolker bbolker at gmail.com
Sat Nov 17 22:09:00 CET 2012

Adam Smith <raptorbio at ...> writes:

> Sorry all, here's a plain text version that is hopefully more user-friendly...
> Original post follows:
> All,
> Can someone please clarify for me why lme4 (glmer) and glmmADMB
> would generate essentially identical estimate of fixed and random
> effects in a Poisson log-normal (individual-level random effect for
> overdispersion) but give (often dramatically) different residual
> values?  It must have something to do with the way fitted values are
> estimated or SDs are calculated for each observation, yes? 

  This was a bug in glmmADMB (forgot to scale the random effects
values by their standard deviations when computing fitted values).
Fixed on R-forge version (0.7.3) now. Thanks!

  Ben Bolker

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