[ADMB Users] gradient_structure problem with nbinom glmm.admb in R
mbrooks at ufl.edu
Mon Jun 13 19:59:46 PDT 2011
If that's what you want, then you are miss specifying the model and changing the max nvar offset won't help.
In lme4 you would specify a random effect as (effect|group) in your case an intercept is an effect of 1 (constant) and you want nested grouping variables. For a quick reference see http://glmm.wikidot.com/faq and scroll down to model specification.
The old version of glmm.admb (the one you're running) can't handle multiple grouping variables, but see http://glmmadmb.r-forge.r-project.org/
You'll have to switch to the new version to get the model you want and an added bonus is that the syntax is more like lme4 which you're used to.
On 12 Jun 2011, at 8:11 AM, Rafael Mares wrote:
> Hi Mollie
> Thanks for the quick reply. I'm trying to run an intercept only model
> with individual identity (group="id") and group identity
> (random=~group) as random effects. I have 230 individuals and 15
> groups (14 - 200 individuals per group, and 1 - 60 measures per
> individual). I usually use lme4, where I would have individuals nested
> in groups (as individuals are unique to groups), but I am aware that
> this isn't possible with glmmADMB (or at least not obvious), so I am
> just trying to specify them as two separate random effects (two
> sources of variation for the intercept)... hopefully this makes sense.
> Maybe I'm not specifying this correctly in the model - I'm not yet too
> familiar with the notation in glmm.admb
> I can run the model with ' random=~1, group="id" ' with no problem,
> but that leaves my group identities out of the picture. Any
> Thanks again for your help and interest.
> Rafael Mares
> Large Animal Research Group (LARG)
> Department of Zoology
> University of Cambridge
> Downing Street
> CB2 3EJ
> On 12 June 2011 12:23, Mollie Brooks <mbrooks at ufl.edu> wrote:
>> Hi Rafael,
>> You might be miss-specifying your model.
>> You may want
>> This would make an intercept only model but allow the different levels
>> of "id" to vary in their intercept. Depending on your experimental design,
>> this might be what you want. What relationship were you hoping the model to
>> Mollie Brooks
>> Ph.D. Candidate
>> NSF IGERT Fellow
>> Biology Department
>> University of Florida
>> mbrooks at ufl.edu
>> On 12 Jun 2011, at 7:04 AM, Rafael Mares wrote:
>> Dear all
>> I am trying to run the following model using glmmADMB (ver. 0.5-2) in
>> R (ver. 2.13.0) in a linux terminal:
>> and I get the following error message:
>> Current maximum number of independent variables is 50000
>> You need to increase the global variable MAX_NVAR_OFFSET to 58531
>> This can be done by putting the line
>> before the declaration of the gradient_structure object.
>> Error in glmm.admb(dur ~ 1, random = ~group, group = "id", data = dat, :
>> The function maximizer failed
>> I've done some reading, mainly on the admb-project website and posts
>> on this mailing list, and tried modifying the 'gradient_structure'
>> line in the TOP_OF_MAIN_SECTION in the nbmm.tpl file in my R library,
>> as I think is suggested... but I get the same error message as if
>> nothing had been changed. I would greatly appreciate it if someone
>> could clarify if I'm doing something wrong, or if it actually isn't
>> possible to change the gradient_structure when running glmmADMB
>> through R. By the way, I have no problems running the same model with
>> Thank you in advance for any help.
>> All the best,
>> Rafael Mares
>> Large Animal Research Group (LARG)
>> Department of Zoology
>> University of Cambridge
>> Downing Street
>> CB2 3EJ
>> Users mailing list
>> Users at admb-project.org
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the Users