[R] Specify order of groups negative binomial (glm.nb)

[Michael Friendly]

On 2/25/2016 1:53 PM, Katharine Miller - NOAA Federal wrote:
>   The dependent
> variable is catch per unit effort (CPUE), and the independent variable is
> the the tributary (Trib_cat).  CPUE is derived from the fish counts divided
> by the effort, so the response is not a count per se, but I think the
> negative binomial is appropriate because of the large numbers of zeros in
> the dependent variable.

You probably shouldn't be dividing by effort -- better to use effort as 
an offset (but I'm not certain that glm.nb supports that)

For the excess zeros, consider using a zero-inflated NB or a hurdle NB
model from the pscl package or countreg package (only on R-Forge)

-Michael