An embedded and charset-unspecified text was scrubbed... Name: not available URL: <https://stat.ethz.ch/pipermail/r-sig-mixed-models/attachments/20130604/3a7bf666/attachment.pl>
issues with weights in glmer (or glmmADMB) [SEC=UNCLASSIFIED]
3 messages · Steve Candy, Ben Bolker
Steve Candy <Steve.Candy at ...> writes:
I am yet to see any response from the developers/maintainers of
lme4 with respect to my posting from 25
Feburary on the R-sig-mixed-models Digest, Vol 74, Issue 40
5. Re: lmer residual variance estimate with prior weights (lmer
wrong, lme and asreml correct) [SEC=UNCLASSIFIED] (Steve Candy)
The regression coefficients from lmer are correct
but the standard errors are way too small in my example
with prior (or "analytical") weights where it
was not appropriate to scale the weights to sum to 1.
As previous postings note, using prior weights of sample
sizes when binomial proportions are the response
variable is a special case but I am not confident to use lmer
with prior weights until the issue my above
posting raises gets some traction.
Sorry this slipped through the cracks. We will take a look; if you can (re)post the query at https://github.com/lme4/lme4/issues?state=open (the new site for bug/feature requests) that would help us out ... cheers Ben Bolke
Ben Bolker <bbolker at ...> writes:
Steve Candy <Steve.Candy <at> ...> writes:
I am yet to see any response from the developers/maintainers of lme4 with respect to my posting from 25 Feburary on the R-sig-mixed-models Digest, Vol 74, Issue 40 5. Re: lmer residual variance estimate with prior weights (lmer
[snip]
Sorry this slipped through the cracks. We will take a look; if you can (re)post the query at https://github.com/lme4/lme4/issues?state=open (the new site for bug/feature requests) that would help us out ...
PS I looked back at your previous post, and you don't provide a reproducible example. If you can either provide your data or (very much preferably) a minimal reproducible example <http://tinyurl.com/reproducible-000>, that will save us time and greatly increase the probability that we will be able to tackle this in the near future ... cheers Ben Bolker