[R-sig-ME] glmmTMB: profile confidence interval scale

Ben Bolker bbo|ker @end|ng |rom gm@||@com
Wed Jun 22 16:11:49 CEST 2022

   This is a documentation bug, will fix.

   If you go to the *current* (online) version of the 'covstruct' 
vignette (which is what was meant to be referred to, and which corrects 
an error found in the current version):  ρ=θ/sqrt(1+θ^2), where θ is the 
Cholesky-scale parameter and rho is the correlation.


On 2022-06-22 6:38 a.m., Aaron Renggli wrote:
> I have an issue using profiled confidence intervals in a random slope
> model.
> The helpfile for confint.glmmTMB regarding the different methods to compute
> confidence intervals says:
> " [...] while "profile" and "uniroot" report them on the underlying
> ("theta") scale: for each random effect, the first set of parameter values
> are standard deviations on the log scale, while remaining parameters
> represent correlations on the scaled Cholesky scale (see the"
> It just cuts off there.
> I don't understand what is meant by the scaled Cholesky scale for the
> correlation parameters. How would I transform the respective confidence
> interval back to the original scale?
> Thank you for your help
> 	[[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

More information about the R-sig-mixed-models mailing list