I have a dataset that is a proportion of home range overlap (HRO) ranging
from 0-0.9. Depending on the dataset there are only a few zeros (~6) or
many (~50). There are multiple years worth of data for most individuals, so
I've included ID as a random effect. From what I've read, a zero-inflated
beta regression model, fitted using a hurdle model is most appropriate. I
realize I can do this in a two-step modelling process, but it also sounds
like glmmTMB is capable of doing this in a one-step hurdle model.
My current model formula (for HRO>0) is:
sf.beta<-glmmTMB(HRO ~ Dist+site+(1|ID), data=NI95,
family=list(family="beta",link="logit"))
I haven't specified zi because I don't think it's necessary until I add the
zeros (though I could be wrong about this). Is it possible to use a hurdle
model for this example? What would the formula look like?
Thanks in advance!