11 November 2015 12 6K Report

I am analyzing the variations of observed species number in a temporal gradient. When I conducted a poisson GLM, I found that the scale parameter was about 0.25. The dispersiontest() function under the AER library also indicates significant underdispersion.

I know I can use quasi-poisson model under the overdispersion condition. When I searched it, I found some documents stated that Quasipoison could also be used underdispersed data, while others did not suggest to use it. 

What I would like to ask is, may I use a quasi-poisson model for strongly underdispersed data. If not, what do you suggest instead?

Similar questions and discussions