Date:         Tue, 28 Oct 2003 09:06:51 -0600
Reply-To:     "Striley, Cathy" <STRILEYC@EPI.WUSTL.EDU>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         "Striley, Cathy" <STRILEYC@EPI.WUSTL.EDU>
Subject:      log likelihoods in negative binomial regression

Thanks for all the help I have received before from this list!

I am exploring an anomaly between SAS and STATA output.

I am using PROC GENMOD with these options dist=nb link=log type3 wald.

Comparing this to STATA nbreg, I find that my dispersion parameters are very
close to the same value, my slope coefficients are close but not exact,  but
my log likelihoods are very, very different  (SAS =-.563.3051;  STATA
-1428.2415).

Does anyone have a possible explanation?

Thank you for any help!
Cathy

Catherine Woodstock Striley, MSW, LCSW, Ph.D.
NIMH Postdoctoral Fellow
Washington University School of Medicine
Department of Psychiatry
40 N. Kingshighway, Suite 4
St. Louis, MO 63108
StrileyC@epi.wustl.edu
314-286-2268
FAX 314-286-2265