Skip to content
Prev 7123 / 20628 Next

hurdle negative binomial models with MCMCglmm

On 12/11/11 1:25 AM, Zelda Van der Waal wrote:
No, MCMCglmm does not fit a negative binomial model (with or without 
zero component).

However, it does fit an "over-dispersed" Poisson, which includes an 
additional, per-observation random-effect.  Although this is not 
identical to a negative binomial, they are often functionally quite 
similar, and the OD Poisson model is almost always a far superior fit a 
straight Poisson.  In fact, MCMCglmm automatically fits this model (the 
"units" term in the output is the over-dispersion random-effect).

If you are truly committed to a zero-inflated negative binomial, the 
glmmADMB package has some facility for fitting these models; however, I 
believe their are some restrictions on how covariates enter the logit 
vs. count portions of the model.  At least, it seemed that way when I 
perused them a while back.  Ben Bolker could speak more to what is (or 
is not) possible via glmmADMB.

Hope that helps.

cheers, Dave

Dave Atkins, PhD
Research Associate Professor
Department of Psychiatry and Behavioral Science
University of Washington
datkins at u.washington.edu

Center for the Study of Health and Risk Behaviors (CSHRB)		
1100 NE 45th Street, Suite 300 	
Seattle, WA  98105 	
206-616-3879 	
http://depts.washington.edu/cshrb/
(Mon-Wed)	

Center for Healthcare Improvement, for Addictions, Mental Illness,
   Medically Vulnerable Populations (CHAMMP)
325 9th Avenue, 2HH-15
Box 359911
Seattle, WA 98104
http://www.chammp.org
(Thurs)