[R-sig-ME] Logistic exposure glmer polynomial covariate issue

[Ben Bolker]

   I worked on this a little more, posted at

https://github.com/bbolker/mixedmodels-misc/blob/master/notes/rsm_logexp.R

   I tried the power-logistic link and got it to work with linear (but 
not quadratic) models: it does appear to do a bit better on this data 
set, although with the small sample I wouldn't take that too seriously. 
(I left the predictor as G250, not log(1+G250).)

   I would certainly try some of this stuff out on your full data set 
(but be careful not to fish/snoop too much ...)

On 5/17/21 1:43 PM, Kirsti Carr wrote:
> Hi Ben,
> 
> Thank you for your thoughts. I will look into cloglog and your other
> suggestions. I should note that the data I provided was not the full
> dataset (after reading the recommendations for posting I decided to provide
> a minimum dataset that still produced the error). Actual effective sample
> size/ number of failures is 23. Still small -- so perhaps taking out the
> cubic and quadratic models is best? I can also provide full dataset if
> necessary.
> 
> Thanks so much,
> Kirsti
> 
> 	[[alternative HTML version deleted]]
> 
> _______________________________________________
> R-sig-mixed-models using r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models
>