From: "Olofsen, E. (ANST)" 
Subject: [NMusers] correlation function for the epsilons
Date:Wed, 14 May 2003 22:35:57 +0200

Dear group,

I wanted to specify a correlation function for the epsilons as discussed in:


http://www.cognigencorp.com/nonmem/nm/98jul101998.html


Furthermore, I wanted to test this with single-subject data. At the end of
the above link it is noted that the single-subject data should be entered
as population data, but with OMEGA constrained to zero. However, when the
data file and $INPUT contain the ID item, the covariance step does not give
correct results (the estimation step does). If I remove `ID', NONMEM gives
a warning about the SIGMA matrix, it treats every observation as coming from
a different individual, and I wonder if the above technique will work.

Does anyone know how I should proceed?

Thanks in advance,

Erik Olofsen
Leiden University Medical Center

PS: An example (t.d contains e.g. [1 0; 1 1; 1 2; 1 3]):

$PROBLEM indeed
$DATA t.d
$INPUT ID DV
$PRED
A = THETA(1)+ETA(1)
Y = A+ERR(1)
$THETA
1
$OMEGA
0 FIX
$SIGMA
0.1
$ESTIMATION MAX=9999 SIG=4 PRINT=1
$COVR 
_______________________________________________________