[R] Error fitting overdispersed logistic regression: package dispmod

[Jason Curole]

Hi all,

First, a quick thank you for R; it's amazing.

I am trying to fit models for a count dataset following the overdispersed logisitic regression approach outlined in Baggerly et al. (BMC Bioinformatics, 5:144; Annotated R code is given at the end of the paper) but R is returning an error with the data below.  Any help in understanding or overcoming this obstacle is appreciated.

>library(dispmod) # required for dispersion fitting

# Now the data

>counts<-matrix(c(2,1,3,1,2715597,3296062,2945864,2215143), ncol=2)
>temp<-c(25,20,25,20) # linear factor-Temperature
>trtmnt<-c(0,0,1,1) #categorical factor

#And the models

>fit1<-glm(counts~temp+trtmnt, family="binomial")
>fit2<-glm.binomial.disp(fit1)

Binomial overdispersed logit model fitting...
Iter.  1  phi: -3.615313e-07 
Error in glm(formula = counts ~ temp + trtmnt, family = "binomial", weights = disp.weights) : 
  negative weights not allowed

> disp.weights
[1]  54.865991  -5.218398 -15.379204   5.021180

So, clearly some dispersion weights are negative, which, according to my understanding of the model, would produce negative variances.  Is there a way around this?  

Thanks, Jason