[NMusers] NONMEM V 1.1 and VI 1.0 Bug Alert

NONMEM V 1.1 and VI 1.0 Bug Alert
 
NMTRAN Code
 
2. Background: There is a bug in NMTRAN. This has been present since the
earliest
version of NMTRAN, and applies to NONMEM III, NONMEM IV, NONMEM V, and
NONMEM VI.
 
Sometimes, a blank record is present at the end of the NONMEM data set.
This is read by NONMEM as a record with ID=0 and MDV=0 and DV=0.
Assuming that the final record of the input data set has an ID value
that is not 0 (which is highly likely), then there is one extra
individual record and one extra observation record. This will affect
the results of the Estimation step.
 
The bug can affect problems with the following characteristics.
 
1) The data set has more than 9999 records (so that a NONMEM FINISH
record is generated at the end of the the NONMEM data set FDATA).
 
2) NMTRAN does not append EVID, MDV, or ID data items. This happens
when the problem uses $PRED, or uses PREDPP and all of EVID, ID, MDV
are listed in $INPUT. (Note that when the data are single subject,
NMTRAN appends the ID data item unless $INPUT specifies ID=L1. This
applies to both $PRED and PREDPP.)
 
3) NM-TRAN generates a multi line format for the NONMEM data set FDATA,
i.e., the FORMAT in FCON contains "/", and the number of lines in FDATA
is at least twice the number of lines in the input data set + 1.
 
4) Let w = the maximum width of any data character + 1
         (i.e., the maximum number of characters in any data item + 1)
   Let m = the number of data items listed in $INPUT, less any that
           are DROPed
   Certain values of w and m can give rise to the bug. These are:
    w m
   10 15
   11 13
   11 19
   11 20
   13 11
   13 16
   13 17
 
   An example of a data item for which w=10 is 0.0098269
 
The user should always examine the following lines in the NONMEM output
report.
 TOT. NO. OF DATA RECS:
 TOT. NO. OF OBS RECS:
 TOT. NO. OF INDIVIDUALS:
 
If the values of OBS RECS and INDIVIDUALS are one more than is
expected, then the bug has occurred. (The value of DATA RECS will be
incorrect also, but the user cannot easily check this.)
 
Workaround: For NONMEM VI, there is a workaround.
Include the option WIDE on the $DATA record.
(With WIDE, there is no FINISH record).
 
For earlier versions of NMTRAN and NONMEM, there is no workaround.
 
Fix: There is a fix to NMTRAN's READ3 subroutine. It applies to all
versions
of NMTRAN and NONMEM.
 
In READ3
 
Locate the line:
      KF=INT(LFMAX/W)
 
Insert after it:
C ADDED 1/2007
      IF (W*(M-KI*KF).GT.LFMAX) KI=INT(M/KF)



___________________________________________________________________________=
____________________________________________________________

  

 
This e-mail transmission may contain confidential or legally privileged =
information that is intended only for the individual or entity named in the=
 =
e-mail address. If you
are not the intended recipient, you are hereby notified that any disclosure=
,=
 copying, distribution, or reliance upon the contents of this e-mail is =
strictly prohibited.
If you have received this e-mail transmission in error, please reply to the=
 =
sender, so that ICON can arrange for proper delivery, and then please
delete the message. Thank You.



 









Received on Wed Jan 17 2007 - 14:12:01 EST