R help - Sep 2009 - Question about zero-inflated poisson with REML.

Dear All,

As you know, glmmADMB package use ML method for estimation.
Is it possible to use REML estimation method for zero-inflated Poisson
distribution?

For ML method,
poi_ML <- glmm.admb(los ~ psihigh + trt.mod + trt.high + psihigh*trt.mod +
psihigh*trt.high + 1, random = ~1, group="site",
family="poisson",
data=edcap)
summary(poi_ML)

How can I control to use REML method estimation in glmm.admb?
Thank you.

Sincerely,
SK

	[[alternative HTML version deleted]]

Sunghee Kim <sunghee0701 <at> gmail.com> writes:
> 
> Dear All,
> 
> As you know, glmmADMB package use ML method for estimation.
> Is it possible to use REML estimation method for zero-inflated Poisson
> distribution?
> 
> For ML method,
> poi_ML <- glmm.admb(los ~ psihigh + trt.mod + trt.high + psihigh*trt.mod
+
> psihigh*trt.high + 1, random = ~1, group="site",
family="poisson",
> data=edcap)
> summary(poi_ML)
> 
> How can I control to use REML method estimation in glmm.admb?
> Thank you.
  I don't think you can do this within glmm.admb, but ADMB is now free
& open source, and information on how to do this (although not
trivial) is in the ADMB-RE manual.  See admb-project.org ...