[R-sig-ME] How to know if random intercepts and slopes are necessary for glmer.nb model

[Ben Bolker]

Some relatively stream-of-consciousness thoughts:

* what kind of errors?  glmer.nb is still not as robust as we (or
you!) would like, e.g. see <https://github.com/lme4/lme4/issues/319>
... information on the kinds of warnings & errors you're getting would
be useful.
* in general I would say that you should *try* to keep the random
effects in if you can -- it is a bit of a catch-22 if you can't fit
the models though ...
* do you get similar results for simulated data with similar structure?
* you could try alternative fitting platforms _or_ alternative models
to account for dispersion, specifically
    * glmmADMB (could be slow for large data sets?)
    * glmmTMB (experimental! <https://github.com/glmmTMB/glmmTMB> )
    (any other suggestions welcome ...)
    * using an observation-level random effect rather than NB to
account for dispersion (in my experience these tend to give similar
results: Harrison 2015 <https://peerj.com/articles/1114/> does a
simulation study for the analogous case of overdispersed binomial
models and concludes "use with caution ...")

On Mon, Oct 19, 2015 at 8:59 AM, David Jones <david.tn.jones at gmail.com> wrote:
> I am receiving a number of different warnings/errors when running glmer.nb
> on a fairly large dataset (N>500,000). For some of the models I have run,
> program-reported errors prevent the generation of estimates. I suspect that
> it is because the random effects are very small. I have tried models with
> random intercepts, as well as models with both random intercepts and slopes
> (all models include fixed effects). I am running models on a dataset which
> in theory would include random effects (patients nested within hospitals).
>
> My question is: how do you know if random intercepts and slopes are
> necessary, if you can't even estimate the random effects models (and thus
> use a model comparison test)? As I am aware you can look at design effects
> to evaluate if a random intercept is necessary (though please correct me if
> I am wrong here).
>
> Some example code I have used is below - many thanks.
>
> a2 <- as.factor(analysis$Location)
> NBIntercept<- glmer.nb(y ~ a2 + (1 | Hospital), data = analysis)
> NBInterceptSlope <- glmer.nb(y ~ a2 + (1 | Hospital) + (1 + a2 | Hospital),
> data = analysis)
>
>         [[alternative HTML version deleted]]
>
> _______________________________________________
> R-sig-mixed-models at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models