1

I am unsure whether turning count data into density still makes Poisson as the appropriate distribution for modelling.

If a given response data are, for example, counts of animals, then such data are best modelled using a Poisson (or quasi-Poisson) distribution.

If the response data are continuous (but bound by zero), then the gamma distribution is likely appropriate.

But what happens when data from counts are turned into density (number of animals per unit area) or encounter rate (number of animals per unit of effort)? Is Poisson distribution still appropriate? Or would one need to use a gamma (or some other) distribution?

I assume it should still be Poisson (or quasi-Poisson, in case of overdispersion).

Tilen
  • 740
  • 7
  • 18
  • See also [In a Poisson model, what is the difference between using time as a covariate or an offset?](http://stats.stackexchange.com/q/175349/17230) & [When to use an offset in a Poisson regression?](http://stats.stackexchange.com/q/11182/17230). – Scortchi - Reinstate Monica Jan 13 '17 at 14:25
  • Thank you @Scortchi. While I looked for similar topics, this one escaped me. Many thanks! – Tilen Jan 13 '17 at 15:17
  • You're welcome! Sometimes it's just a matter of happening on the right key-word - "offset" in this case. Please up-vote any questions/answers you find helpful. – Scortchi - Reinstate Monica Jan 13 '17 at 15:43

0 Answers0