From: "Ludden, Thomas (MYD)" luddent@iconus.com
Subject: [NMusers] Epsilons correlated across L2 records
Date: 12/4/2003 9:49 AM

Professor Stuart Beal has asked that the following message be posted to
nmusers.

Tom Ludden
GloboMax
The Strategic Pharmaceutical Development Division of ICON plc

______________________________________________________________________

From: Stu Beal
Subject: [NMusers] Epsilons correlated across L2 records
Date: December 3 2003

Few NONMEM analyses  have  used  correlated  epsilons  across  L2  records.

Regrettably  though, we have now learned that the results from any previous

attempt to do this (using  NONMEM-PRED  common  NMPRD5)  *with  conditional

estimation* are erroneous.


If the user wishes to use correlated epsilons across L2 records with condi-

tional  estimation,  then  the following code change in NONMEM routine OBJ2

should be made.


LOCATE the two lines of code:

         IQ=0
         IF (MODE.EQ.2) THEN

REPLACE these with:

         IF (MODE.NE.3) THEN
         IQ=0

Once OBJ2 source code has been changed, it must be recompiled.  (See  Users

Guide  III,  p,  125 and p. 129-130, in which we suggest erasing all object

modules (.obj or .o) in the nm subdirectory and then using the  SETUP  com-

mand as it was originally typed.)
_______________________________________________________