Hello R-list. I am a "long time listener - first time caller" who has
been using R in research and graduate teaching for over 5 years. I
hope that my question is simple but not too foolish. I've looked
through the FAQ and searched the R site mail list with some close hits
but no direct answers, so...
I would like to estimate QAIC (and QAICc) for a glm fit using the
quasibinomial family. I found a general reference suggesting a simple
solution:
"we calculated QAICc adjusting for overdispersion by dividing the
residual deviance (i.e. -2 loglikelihood) with the overdispersion
parameter calculated from the most complex model as the sum of squares
Pearson residuals divided by the number of degrees of freedom (Burnham
& Anderson, 2002). "
- Mystrud et al. 2007. Animal Conservation. 10:77-87.
My question is: Will this calculation be valid with the residual
deviance returned by the glm() function using the quasibinomial family
as reported in R?
I thought I should ask to be certain that there is no dispersion
correction applied to the reported deviance, as encouraged by Burnham
and Anderson, 2nd ed., 2002 on p.69:
"When data are overdispersed and c > 1, the proper likelihood is log(L)/c".
Regards, Darren Gillis
Department of Biological Sciences
Faculty of Science
University of Manitoba
Winnipeg, MB
Canada, R3T 2N2
______________________________________________
R-help@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
