Date:         Fri, 8 Sep 2000 17:34:29 -0700
Reply-To:     Cassell.David@EPAMAIL.EPA.GOV
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         "David L. Cassell" <Cassell.David@EPAMAIL.EPA.GOV>
Subject:      Re: Poisson Vs Negative Binomial Regressions
Comments: To: nwohaj@spectre.ag.uiuc.edu
Content-type: text/plain; charset=us-ascii

Ogbonnaya John Nwoha wrote:
> How can I determine the better of a poisson regression result and a
> negative binomial regression result.  Attached are the outputs of an
> actual regression.  Which one is better?

I would hesitate to judge "better" based on mere statistical performance,
particularly when your test statistics indicate that neither fit is really a lot
better than the other.  [I did have real trouble reading your attachments so
I could have overlooked something - in future please do not use MIME
in a mailing list which is gateway'd to a Usenet newsgroup!]

You'll be better off attacking this problem from a different direction.  Or two.
Here are two to start with.  First, how do the residual plots look?  Is one
clearly a better fit than the other?  Is one clearly meeting the underlying
assumptions of the modeling process better than the other?  And second
[this is where I would actually start if I were consulting on this], what are
the
underlying features of your process?  How do you expect the process to
work, and hence which of the two models is a more rational model from
the *scientific* perspective?

"Don't let statistics drive your science; let the science drive your
statistics."

David
--
David Cassell, OAO Corp.                           Cassell.David@epa.gov
Senior computing specialist
mathematical statistician