st: Exact Poisson regression

[Garry Anderson <g.anderson@unimelb.edu.au>]

Hi Statalisters,

 

Is there a method that can use exact Poisson regression when one assumes that the distribution is a negative binomial? I have a small dataset (n=21) of counts and analyses with -poisson- and -nbreg- indicate that alpha in nbreg is greater than zero (chibar2(01) =   13.21, P<0.001). I suppose an exact negative binomial regression would be preferable, but I doubt if it has been implemented.

 

Any thoughts would be much appreciated.

 

Regards, Garry