Date:         Thu, 22 Dec 2011 08:05:02 -0500
Reply-To:     William Shakespeare <shakespeare_1040@HOTMAIL.COM>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         William Shakespeare <shakespeare_1040@HOTMAIL.COM>
Subject:      Proc glimmix

I have a two level model that has subjects in locations and a binary
outcome.  I want to treat locations as random so each location has it's
own regression line and all have the same slope.  A1 is continuous with
variance at level 1 only, a2 is continuous with variance at level 1 and
level 2 and a3 is categorical.

I fit a random intercept model with proc glimmix:

proc glimmix data=data1;
  class a1 a2 a3 loc id;
  model outcome=a1 a2 a3;
  random loc/subject=id;
run;

I'm not sure this accounts for the level 2 variance in a2 or constrains
the level 2 variance in a1 to zero.  Suggestions/comments?