[R-sig-ME] Model Validation with glmmADMB
Mondain-Monval, Thomas
t.mondain-monval at lancaster.ac.uk
Thu Nov 5 17:23:24 CET 2015
Hi,
I am pretty new to mixed effects models and am using glmmADMB to analyse some bird reproductive success data. I have decided to use a hurdle model as I have a lot of true zeros in my dataset. I'm using the glmmadmb function with "truncnbinom2" as the family to analyse the non-zero part of the data first. However, now that I have my model and want to do validate it, I have run into a problem. When I call the residuals for the model all I get is a column of NAs with no actual values. All the model validation techniques I have previously used require residuals to do them, e.g. residuals vs fitted, residuals vs explanatory variables etc. What other validation techniques can you suggest? Or is there a way of getting residuals out of the model?
This is my full model:
non0ri<-glmmadmb(fledging.number~scnest.depth*scbrood.number + (1 | colony/burrow.code),
family="truncnbinom2",
data=subset(no.na, fledging.number>0))
Where fledging number is subsetted to only include non-zero data.
And I used residuals(non0ri) to try and get the residuals.
I hope you can help.
Thank you and best wishes,
Thomas
[[alternative HTML version deleted]]
More information about the R-sig-mixed-models
mailing list