[R] Checking modeling assumptions in a binomial GLMM
Ben Bolker
bbolker at gmail.com
Wed Jul 16 22:01:02 CEST 2014
Ravi Varadhan <ravi.varadhan <at> jhu.edu> writes:
>
> Dear All,
> I am fitting a model for a binary response variable measured
> repeatedly at multiple visits. I am using the binomial GLMM using
> the glmer() function in lme4 package. How can I evaluate the model
> assumptions (e.g., residual diagnostics, adequacy of random effects
> distribution) for a binomial GLMM? Are there any standard checks
> that are commonly done? Are there any pedagogical examples or data
> sets where model assumptions have been examined for binomial GLMMs?
> Any suggestions/guidance is appreciated.
>
> Thank you,
> Ravi
This might be better for r-sig-mixed-models at r-project.org.
Roughly speaking, you want to do one set of diagnostics on
the individual-level residuals similar to those for a binomial GLM
(which in turn are adaptations of the diagnostics for linear models)
and one on the group-level random effects. As with GLMs, if your
binomial values are _binary_ then the individual-level diagnostics
will be a bit challenging. Binomial GLMMs with N>1 will be a bit
easier.
http://rpubs.com/bbolker/glmmchapter may be helpful, especially the second
("Culcita") example.
Also http://stats.stackexchange.com/questions/70783/
how-to-assess-the-fit-of-a-binomial-glmm-fitted-with-lme4-1-0/
(broken URL to make Gmane happy)
More information about the R-help
mailing list