[R-sig-ME] binomial GLMM with small upper limit

David Duffy David.Duffy at qimr.edu.au
Thu Jun 21 01:00:03 CEST 2012


On Wed, 20 Jun 2012, Wojdak, Jeremy wrote:

> I am working with mortality data - the number of three animals per 
> experimental unit that had died by week 1, week2, ... during an 
> experiment.  So, I am using a binomial GLMM approach to model 
> proportions with a random effect to deal with the repeated measurements 
> from the same exp. unit.  (Specifically, I use the experimental unit or 
> "Tank" as the random effect, since multiple observations from the same 
> unit must be related... and there were temporal/spatial blocks, so each 
> tank is nested within a block.  I include sample "Day" as a fixed effect 
> in the model)
>
> sm7<-glmer(predc.survtbl~predator*Day +(1|Block/Tank), 
> family="binomial", data=predc2)

Does your riskset size change by Day in predc.survtbl?

> All is well, except the model suggests there are no treatment or time 
> effects, while graphical inspection suggests there may be both.

So a naive survival analysis (ignoring Tank etc) is significant?  You
might try the coxme package...


-- 
| David Duffy (MBBS PhD)                                         ,-_|\
| email: davidD at qimr.edu.au  ph: INT+61+7+3362-0217 fax: -0101  /     *
| Epidemiology Unit, Queensland Institute of Medical Research   \_,-._/
| 300 Herston Rd, Brisbane, Queensland 4029, Australia  GPG 4D0B994A v



More information about the R-sig-mixed-models mailing list