Date:         Thu, 26 Apr 2007 16:30:39 -0400
Reply-To:     Mike Todd <mtodd@PREV.ORG>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         Mike Todd <mtodd@PREV.ORG>
Subject:      Multilevel poisson with overdispersion in NLMIXED?

Hi all:

I'm trying to run an analysis in NLMIXED that parallels that specified by
the following GLIMMIX code:


proc glimmix data=a noclprint;
  class sch_id stud_id;
  model y=x
   /s ddfm=bw dist=poisson;
  random intercept/sub=sch_id;
  random _residual_;
run;

In this simple model, we have students nested within schools, and we are
predicting (student-level) y from (student-level) x, and specifying a random
component for school-level intercept and extra-Poisson variability in the
student-level residuals.

Is such a specification possible in NLMIXED? If so, how would I go about
setting such a model up?

Thanks for any help you can provide.

-Mike