[R-sig-ME] Error in lme4 0.999902344-0: "Object 'multResp' not found" ?
Ben Bolker
bbolker at gmail.com
Fri May 25 16:40:44 CEST 2012
Bertolt Meyer <bmeyer at ...> writes:
>
> Ben Bolker <bbolker at ...> writes:
>
> > thanks for the notification. I've just posted 0.999902345-0
> > to the repository (it may take up to 24 hours to appear), should
> > address this issue.
>
> That does indeed solve the problem, thanks for the quick fix!
> A quick question: A colleague told me that this
> newest version of lme4 can do CIs for the parameter estimates of
> the fixed effects. Does that require
> bootMer() or is there another way?
In principle yes; profiling a merMod (new-lme4) object creates
a profile object of class "thpr", which has a confint method.
The confidence intervals produced in this way are more accurate than
the Wald intervals implied by summary() (analogous to the difference
between confint.default() and confint.glm() applied to a glm() fit),
but are still based on Z-score/likelihood ratio test cutoffs, rather
than trying to do any kind of finite-size correction. If you want
more accurate confidence intervals that incorporate the effects of
the sample size, then you'll need bootMer.
However ... for GLMMs, we are still wrestling a bit with the
robustness of the fitting, which tends to be more of a problem
when profiling -- so it may not be practical to get profile
confidence intervals on GLMMs quite yet ...
Ben Bolker
More information about the R-sig-mixed-models
mailing list