[R-sig-ME] glmer and influence.me - complaining about nAGQ==0

[Ben Bolker]

On 5/3/21 1:20 PM, Cátia Ferreira De Oliveira via R-sig-mixed-models wrote:
> Dear Professor Bolker,
> 
> Thank you for your comment. I haven't been able to find this information
> anywhere, but just wanted to check something you said, is the "(tol =
> 0.002, component 1)" the threshold you referred to?

   Yes.  I was comparing the max|grad| value of 0.00209 with the 
tolerance of 0.002.  This tolerance can be adjusted in ?glmerControl 
(e.g. glmerControl(check.conv.grad = .makeCC(tol=3e-3)) )

> 
> optimizer (bobyqa) convergence code: 0 (OK)Model failed to converge with
> max|grad| = 0.00209134 (tol = 0.002, component 1)
> 
> Also, if I run the model without the logRT I will still need to use
> the nAGQ==0 right?

    I don't know.  Hopefully it will work better and you'll be able to 
use the default Laplace approx (nAGQ=1) setting

> 
> Best wishes,
> 
> Catia
> 
> On Tue, 27 Apr 2021 at 11:01, <r-sig-mixed-models-request using r-project.org>
> wrote:
> 
>> Send R-sig-mixed-models mailing list submissions to
>>          r-sig-mixed-models using r-project.org
>>
>> To subscribe or unsubscribe via the World Wide Web, visit
>>          https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models
>> or, via email, send a message with subject or body 'help' to
>>          r-sig-mixed-models-request using r-project.org
>>
>> You can reach the person managing the list at
>>          r-sig-mixed-models-owner using r-project.org
>>
>> When replying, please edit your Subject line so it is more specific
>> than "Re: Contents of R-sig-mixed-models digest..."
>>
>>
>> Today's Topics:
>>
>>     1. Re: glmer and influence.me - complaining about nAGQ==0
>>        (=?UTF-8?Q?C=C3=A1tia_Ferreira_De_Oliveira?=)
>>     2. Re: glmer and influence.me - complaining about nAGQ==0
>>        (Ben Bolker)
>>
>> ----------------------------------------------------------------------
>>
>> Message: 1
>> Date: Mon, 26 Apr 2021 17:24:50 +0100
>> From: =?UTF-8?Q?C=C3=A1tia_Ferreira_De_Oliveira?= <cmfo500 using york.ac.uk>
>> To: r-sig-mixed-models using r-project.org
>> Subject: Re: [R-sig-ME] glmer and influence.me - complaining about
>>          nAGQ==0
>> Message-ID:
>>          <CACw+Tfd_s9HKfBNushVtmhWg2=
>> i14yf0QJF1MWuNg8MAAuSgwA using mail.gmail.com>
>> Content-Type: text/plain; charset="utf-8"
>>
>> Thank you for your replies!
>>
>> Regarding your comment about having logRT in a gamma model with log link, I
>> decided to try it after getting this warning if I only use RT as the
>> dependent variable:
>> (glmer(RT ~ ...)
>>
>>
>>
>>
>>
>> *optimizer (bobyqa) convergence code: 0 (OK)Model failed to converge with
>> max|grad| = 0.00209134 (tol = 0.002, component 1)Model is nearly
>> unidentifiable: very large eigenvalue - Rescale variables?*
>>
>> Do you have a better suggestion for dealing with this that does not require
>> the log transformation and that may allow me to use the influence.me
>> package?
>>
>> Best wishes,
>>
>> Catia
>>
>>          [[alternative HTML version deleted]]
>>
>>
>>
>>
>> ------------------------------
>>
>> Message: 2
>> Date: Mon, 26 Apr 2021 13:55:11 -0400
>> From: Ben Bolker <bbolker using gmail.com>
>> To: r-sig-mixed-models using r-project.org
>> Subject: Re: [R-sig-ME] glmer and influence.me - complaining about
>>          nAGQ==0
>> Message-ID: <b515aefb-0bff-d27d-75eb-ff12f4069e42 using gmail.com>
>> Content-Type: text/plain; charset="utf-8"; Format="flowed"
>>
>>
>>     These convergence warnings are not necessarily problematic (see
>> ?lme4::convergence, for example). In particular, the overly large max
>> |grad| is only slightly above the threshold (and, these computations can
>> be *less* reliable for very large data sets); the large eigenvalue is
>> similarly just a warning, not necessarily a problem.
>>
>>     Do model diagnostics (e.g. with DHARMa) generally look OK?  You can
>> try allFit() if you have some patience.
>>
>>     The main thing I would do is think carefully/inspect model
>> predictions to see whether you think RT is the more appropriate scale.
>>
>>
>>
>> On 4/26/21 12:24 PM, Cátia Ferreira De Oliveira via R-sig-mixed-models
>> wrote:
>>> Thank you for your replies!
>>>
>>> Regarding your comment about having logRT in a gamma model with log
>> link, I
>>> decided to try it after getting this warning if I only use RT as the
>>> dependent variable:
>>> (glmer(RT ~ ...)
>>>
>>>
>>>
>>>
>>>
>>> *optimizer (bobyqa) convergence code: 0 (OK)Model failed to converge with
>>> max|grad| = 0.00209134 (tol = 0.002, component 1)Model is nearly
>>> unidentifiable: very large eigenvalue - Rescale variables?*
>>>
>>> Do you have a better suggestion for dealing with this that does not
>> require
>>> the log transformation and that may allow me to use the influence.me
>>> package?
>>>
>>> Best wishes,
>>>
>>> Catia
>>>
>>>        [[alternative HTML version deleted]]
>>>
>>> _______________________________________________
>>> R-sig-mixed-models using r-project.org mailing list
>>> https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models
>>>
>>
>>
>>
>>
>> ------------------------------
>>
>> Subject: Digest Footer
>>
>> _______________________________________________
>> R-sig-mixed-models mailing list
>> R-sig-mixed-models using r-project.org
>> https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models
>>
>>
>> ------------------------------
>>
>> End of R-sig-mixed-models Digest, Vol 172, Issue 27
>> ***************************************************
> 
> 
>