[R-sig-ME] MCMCglmm warning: observations with zero weight not used for calculating dispersion

[Jarrod Hadfield]

Hi,

This is warning in from glm which MCMCglmm uses to get starting  
values. It is not necessarily anything to worry about, but it may  
indicate that the data are set up wrong. I think the most likely  
reason is that Suc_5+Fa_5 = 0 for some observations.

Cheers,

Jarrod


Quoting Julie Rushmore <julierushmore at gmail.com> on Thu, 5 Apr 2012  
10:27:53 -0400:

> Hi all,
>
> I am modeling association patterns among a community of chimpanzees using a
> logistic mixed effects model in the MCMCglmm package. When I run my model,
> I get the following warning message:
>
> Warning message:
> In summary.glm(glm(cbind(MCMC_y, MCMC_y.additional) ~ 1, family =
> "quasibinomial",  :
>   observations with zero weight not used for calculating dispersion
>
> I can't tell if this warning is related to zero-inflation, overdispersion,
> or something else. Also, I'm not sure 1) if this is a serious concern, or
> 2) how to deal with this in the MCMCglmm framework.
>
> Below is the code for my model:
>
> #priors:
> priors1<- list(R = list(V = 1, nu = 0.002), G = list(G1 = list(V =1, nu =
> 0.002), G2 = list(V =1, nu = 0.002)))
> #model:
> m1<- MCMCglmm(cbind(Suc_5,Fa_5)~
> Related+Sex.Estrus+FamilyDiff+RankDiffNew+EstrusAd3*Age, random =
> ~idv(mult.memb(~Actor1+Actor2))+PID, family="multinomial2", data=dat,
> verbose = FALSE, prior=priors1, saveX=TRUE, saveZ=TRUE, pr=TRUE)
>
> *The response variable uses Suc_5 (two IDs observed together) and Fa_5 (two
> IDs not observed together) to calculate pairwise association indices.
> *The fixed effect predictor variables are all categorical and measure
> pair-wise attributes about two individuals in a pair (with the exception of
> EstrusAd3, which is a temporal variable - the number of females in
> estrus/month).
> *Random effects include animal ID (i.e., the multi-member component) and
> pair ID (PID).
>
> Thanks for your feedback!
> Julie
>
> --
> Julie Rushmore
> DVM/PhD Candidate
> Odum School of Ecology
> University of Georgia
>
> 	[[alternative HTML version deleted]]
>
> _______________________________________________
> R-sig-mixed-models at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models
>
>



-- 
The University of Edinburgh is a charitable body, registered in
Scotland, with registration number SC005336.