glmmADMB v 0.7.1

Thu, Jan 12, 2012 9:04 AM

Hi Chris,
You don't need to code categorical predictors with dummy variables. As long as they are factors, the function knows what to do (same for glmer and most others). I don't know if that would cause a problem, but it's possible that it could be wasting space. If that doesn't solve the problem, try writing to the mixed models list.
Mollie

On 12 Jan 2012, at 11:54 AM, Christopher Rota wrote:

Hi Mollie,

The variables 'BrnLight' through 'RedHit' represent categorical variables describing tree condition.  They are coded as dummy variables, and one tree condition category (Green) is omitted from model specification and interpreted as the intercept.  The season (Autumn, Spring, Winter) and year (Yr01, etc.) variables are coded in an identical manner.  DBH and Start.Time are of course continuous variables.

On 1/12/2012 10:44 AM, Mollie Brooks wrote:

On 12 Jan 2012, at 11:31 AM, Christopher Rota wrote:

Hi Mollie,

Here is my code:

fit<- glmmadmb(formula=count~BrnLight + BrnMod + BrnMPB + BrnSev + GrnHit + GryHit + RedHit + Autumn + Spring + Winter + Yr01 + Yr12 + Yr23 + Yr34 + Yr45 + Est.DBH.in + Start.Time + (1|Color.Combo) + (1|Observers), data=beh.data, family='nbinom')

This looks ok, but I wonder, did you separate out levels of the categorical fixed effects? i.e. are BrnLight + BrnMod + BrnMPB + BrnSev levels of Brn, GrnHit + GryHit + RedHit  levels of Hit,  Autumn + Spring + Winter levels of season, and Yr01 + Yr12 + Yr23 + Yr34 + Yr45 levels of Yr ?

Thanks for the prompt response,

Chris

On 1/12/2012 10:27 AM, Mollie Brooks wrote:

Hi Chris,
I am using this function on a dataset larger than this, so I don't think that's the problem, but I could be wrong.
Often, this error indicates that the model was mis-specified. Could you send the call to glmmadmb?
thanks,
Mollie

Mollie Brooks
Ph.D. Candidate
NSF IGERT Fellow
Biology Department
University of Florida
mbrooks at ufl.edu
http://people.biology.ufl.edu/mbrooks




On 12 Jan 2012, at 11:18 AM, Christopher Rota wrote:

Dear R Users,

I am running into some trouble when using glmmADMB version 0.7.1 and am hoping someone in the R community may have some insight.

I am trying to fit a rather large negative binomial mixed-effects model. I have 3980 observations (counts of foraging attempts). My 'global' model has 17 fixed effects and 2 random effects. Fixed effects consist of both continuous and categorical variables. Each categorical variable has at least 22 observations, most have considerably more. One random effect is an 'observer' effect consisting of 11 different observers. Each observer made at least 29 observations, but most made considerably more. The other random effect in an 'individual bird' effect consisting of 78 individual birds. There are at least 20 observations made on each bird.

This global model runs for about 10 minutes, then fails with the following message:
Memory allocation error -- Perhaps you are trying to allocate too much memory in your program

When I monitor my computer performance with Windows Task Manager while the model is running, I can watch Physical Memory Usage slowly tick up. All of that increased memory use is attributed to glmmadmb.exe. I will watch memory use for this program tick up to about 1.7GB, and that is when the model fails. I am using a computer with a Windows Vista 32-bit operating system with 4GB RAM. Is the problem simply that I do not have enough memory on my computer to run this model? If indeed the problem is a shortage of memory, is there any way to make glmmadmb.exe use memory differently, or do I need to use a more powerful computer?

Thank you for any insight.

Chris Rota

--
Christopher Rota
Ph.D. Student

University of Missouri
Fisheries and Wildlife Science
302 Anheuser-Busch Natural Resources Building
Columbia, MO 65211

Office: 303O Anheuser-Busch Natural Resources Buildling
Email: ctr4g2 at mail.missouri.edu
Phone: 573-239-6975
Website: http://www.biosci.missouri.edu/avianecology/rota/index.html
Calendar: http://www.google.com/calendar/embed?src=christopher.rota%40gmail.com&ctz=America/Chicago

_______________________________________________
R-sig-ecology mailing list
R-sig-ecology at r-project.org
https://stat.ethz.ch/mailman/listinfo/r-sig-ecology

glmmADMB v 0.7.1

Thread (8 messages)