[R] AIC for tweedie glm

eleadbeater e.leadbeater at sussex.ac.uk
Thu Sep 30 16:06:25 CEST 2010


Dear R-users,

I'm trying to model some data using a tweedie GLM approach. My response
variable is the number of pupae that are the offspring of a subordinate wasp
on a wasp's nest. However, they're not count data- for each nest, I only
know the mean number of pupae per subordinate, which is continous. The data
also contain a high proportion of zeros.

I'm not very experienced at statistical modelling, but from reading previous
posts, it seems that my data would suit a tweedie approach. I can't use a
zero-inflated Poisson model, because my data are not counts. Many of my
values are between 0 and 1, so if I rounded to the nearest integer, I'd lose
a lot of the variation.

Here's my code:
out<-tweedie.profile(PUPAE_PER_SUB~1,p.vec=seq(1.1,1.9,length=9),method="interpolation",do.ci=TRUE,do.smooth=TRUE,do.plot=TRUE)
tweedie1<-glm(GSA_TOTAL_DF_PERSUB~GROUP_SIZE+PERIOD+SITE+PERIOD*GROUP_SIZE,family=tweedie(var.power=out$p.max,link.power=0))

This worked fine, and gave results I expected, but I don't know what the
best method is to evaluate the fit of the model. I am used to using AIC to
compare models. A site search turned up AICtweedie, within the tweedie
package, but I get the following message: Error: could not find function
"AICtweedie" when I try to use this command, even though "tweedie" and
"statmod" are both loaded. I've also read that AIC can be calculated using
dtweedie, but I'm a beginner and so, despite lots of searching, I'm not sure
how. I'm sorry to ask a basic statistics rather than programming question,
but I'm really stuck. Could anyone advise me on the best way to assess
goodness-of-fit for this type of model, in order to compare models?

Thanks
-- 
View this message in context: http://r.789695.n4.nabble.com/AIC-for-tweedie-glm-tp2720813p2720813.html
Sent from the R help mailing list archive at Nabble.com.



More information about the R-help mailing list