Date:         Fri, 31 May 1996 15:43:00 -0400
Reply-To:     Brian Smith <smith@MSCF.UKY.EDU>
Sender:       "SAS(r) Discussion" <SAS-L@UGA.CC.UGA.EDU>
From:         Brian Smith <smith@MSCF.UKY.EDU>
Organization: University of Kentucky, Dept. of Math Sciences
Subject:      PROC Mixed question

This is one of those questions that is both a statistics
question and a SAS question.  I have two continuous
variables which I hypothesize have a linear relationship.
I have multiple observations per subject, thus it seems
reasonable to use subject as a random effect.  SAS code
follows.
 
proc mixed;
class subject;
model y=x / s;
random subject;
run;
 
When I run this model, I am getting 0 as an estimate for the
variance component.  This seems completely unreasonable and
I believe I might be making a mistake.
 
Statistics question: Under what circumstances would one get
an estimate of 0?
 
SAS question:  When I examine the SAS Log I get the following
messages
 
Note: Estimated G matrix is not positive definite.
 
Note: At least one W.D format was too small for the number to
be printed.  The decimal may be shifted by the "BEST" format.
 
The first note just indicates that the estimate of the variance
component was 0.  What does the second note mean?  Could there
be a relationship between what this note is trying to tell me
and the estimate that I am getting?
 
Sincerely,
 
Brian Smith, PhD