www.phidot.org

I am currently working with a vegetation data set for non-native species in riparian areas in the west. I have a combination of single site and multiple site data across years; within years we also have repeat samples. Given the very unlikely occurrence of non-native vegetation (e.g., cheat grass) being extirpated, I have been fixing Epsilon values to zero. Are there any possible limitations of this approach. I have not found any examples using robust design approaches for non-natives (or vegetation as a response), but it seems very promising given the troubles with detection, increasing trends, etc.

Any information would be great.
Thanks
R0bert