[R-sig-ME] Is there a leverage diagnostic test for glmmTMB models?

[John Fox]

Hi Ben,

Perhaps you, Sandy Weisberg (to whom I'm cc'ing this response), and I 
can sort this out, possibly in the Fall, so that there's more coherent 
functionality for unusual-case diagnostics across a broader range of 
mixed-modeling functions.

Best,
  John

On 2020-07-27 4:25 p.m., Ben Bolker wrote:
>    As I mentioned in my CrossValidated answer 
> https://stats.stackexchange.com/questions/478727/leverage-diagnostic-test-not-supported-for-glmmtmb-models-in-r/479298#479298 
> , this does work with the add-on influence_mixed function in the glmmTMB 
> package ; you just need to specify  group=".cases" (or leave the group 
> blank, since ".cases" is the default value).
> 
>     This doesn't use hatvalues, which as I commented there aren't 
> available (yet) for glmmTMB; instead, it does brute-force leave-one-out 
> computations (although it can do them in parallel).
> 
>     It would be a good idea making this functionality more easily 
> available, but I've messed up in the past when trying to export 'car' 
> functionality, so I'd like to be careful.
> 
> 
>    cheers
> 
>      Ben
> 
> 
> On 7/27/20 3:47 PM, John Fox wrote:
>> Dear April,
>>
>> Unfortunately, there's no hatvalues() method for "glmmTMB" objects, 
>> nor do functions in the car package, where the influenceIndexPlot() 
>> function that you tried to use originates, have "glmmTMB" methods. car 
>> does have methods for the "merMod" objects produced by functions in 
>> the lme4 package and for the "lme" objects produced by the lme() 
>> function in the nlme package, but those probably aren't useful to you.
>>
>> I'm willing to look into providing the necessary infrastructure in the 
>> car package to support glmmTMB() but can't promise to do that before 
>> the Fall.
>>
>> Best,
>>  John
>>
>> --------------------------------------------
>>   John Fox
>>   Professor Emeritus
>>   McMaster University
>>   Hamilton, Ontario, Canada
>>   web: https://socialsciences.mcmaster.ca/jfox/
>>
>> On 7/27/2020 1:40 PM, April Martinig wrote:
>>> Hello,
>>>
>>> I've posted my question to Cross Validated, but I am also linking to 
>>> it here 
>>> <https://stats.stackexchange.com/questions/478727/leverage-diagnostic-test-not-supported-for-glmmtmb-models-in-r> 
>>> because it might have been missed.
>>>
>>> I am trying to use the DHARMa package to test for how much leverage 
>>> specific points (on one extreme) for my response variable has on my 
>>> output.
>>>
>>> In my question I have posted the data, what I have tried, and linked 
>>> to other sources. Thank you in advance.
>>>
>>> Take care,
>>> April
>>>
>>> ---
>>> April Martinig, M.Sc.
>>> Ph.D. Candidate in Ecology
>>> University of Alberta
>>> martinig.weebly.com
>>> “I do not accept the conventional order of things as a given."
>>>
>>> https://stats.stackexchange.com/questions/478727/leverage-diagnostic-test-not-supported-for-glmmtmb-models-in-r 
>>> <https://stats.stackexchange.com/questions/478727/leverage-diagnostic-test-not-supported-for-glmmtmb-models-in-r>. 
>>>
>>>     [[alternative HTML version deleted]]
>>>
>>> _______________________________________________
>>> R-sig-mixed-models using r-project.org mailing list
>>> https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models
>>>
>>
>> _______________________________________________
>> R-sig-mixed-models using r-project.org mailing list
>> https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models
> 
> _______________________________________________
> R-sig-mixed-models using r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models