[R-sig-ME] glmmTMB question - diagnostic plots of model fit and effect sizes
Ben Bolker
bbo|ker @end|ng |rom gm@||@com
Tue Jul 23 20:10:25 CEST 2024
On 2024-07-23 12:27 p.m., Luke D. Fannin wrote:
> Dear Dr. Bolker,
>
> I hope this email finds you well. My name is Luke Fannin; I�m an EEB graduate student at Dartmouth College. I am wondering if I can ask for your advice on some analyses I�ve been running with glmmTMB.
>
> Specifically, I have two questions. Is there a way within the package to: plot a diagnostic plot of model predictions against data to evaluate performance? Also, is there a handy way to calculate parameter effect sizes?
The package itself doesn't have any built-in diagnostic plotting
methods, but you can do something like
plot(fitted(m1), model.response(model.frame(m1)))
If you're willing to use other packages, then
performance::check_model() and the DHARMa packages generally are good
places to start.
I�ve been working with the zero-inflated NB models specifically, but I
wanted to make sure I hadn�t missed any references to these issues in
your helpful GLMM FAQ online. Any help would be most appreciated. Thank
you in advance and I look forward to hearing from you.
This depends very much what you mean by 'effect size'. I often like
the scaled regression coefficients as a low-tech solution (e.g. see the
`by_2sd()` function in the dotwhisker package, or you can scale your
predictor variables yourself). Or take a look around at the helper
packages listed at https://cran.r-project.org/web/views/MixedModels.html
Ben Bolker
>
> Best, Luke
>
> [[alternative HTML version deleted]]
>
>
> _______________________________________________
> R-sig-mixed-models using r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models
--
Dr. Benjamin Bolker
Professor, Mathematics & Statistics and Biology, McMaster University
Director, School of Computational Science and Engineering
(Acting) Graduate chair, Mathematics & Statistics
> E-mail is sent at my convenience; I don't expect replies outside of
working hours.
More information about the R-sig-mixed-models
mailing list