I've notices several times that when
using IMPMAP
and
Running parallel
the "one more iteration for variance" will fail on one (or more) of the
processes. When it does this, the OBJ will be lower than the last iteration of
the
parameter estimation. It seems that when this happens, NONMEM uses the default
value (0) for OBJI for the data sent to that process, resulting in a lower OBJ.
So, you can, in principle, rerun the model using the MSF and single process.
But, interestingly, while the OBJ reported in the outfile files seems to be
wrong, NONMEM will still report SEE etc. Can the SEE values in this case be
believed?
Mark Sale M.D.
Senior Vice President, Pharmacometrics
Nuventra Pharma Sciences, Inc.
2525 Meridian Parkway, Suite 200
Durham, NC 27713
Phone (919)-973-0383
[email protected]<[email protected]>
Different OBJ after covariance with parallel and IMPMAP
Mark:
As mentioned back in August, the failure occurs because your TIMEOUT parameter
in the pnm file was not long enough, and a time out occurred, and NONMEM is not
properly recovering from this. Therefore, the SEE’s should not be believed, as
the subjects from the timed out processes would not be included.
Robert J. Bauer, Ph.D.
Senior Director
Pharmacometrics R&D
ICON Early Phase
820 W. Diamond Avenue
Suite 100
Gaithersburg, MD 20878
Office: (215) 616-6428
Mobile: (925) 286-0769
[email protected]<mailto:[email protected]>
http://www.iconplc.com/
Quoted reply history
From: Robert Bauer [mailto:[email protected]]
Sent: Friday, February 08, 2019 8:08 PM
To: Bauer, Robert
Subject: Fw: [NMusers] Different OBJ after covariance with parallel and IMPMAP
----- Forwarded Message -----
From: Mark Sale <[email protected]>
Cc: [email protected] <[email protected]>
Sent: Monday, February 4, 2019 10:50:28 AM PST
Subject: [NMusers] Different OBJ after covariance with parallel and IMPMAP
I've notices several times that when
using IMPMAP
and
Running parallel
the "one more iteration for variance" will fail on one (or more) of the
processes. When it does this, the OBJ will be lower than the last iteration of
the
parameter estimation. It seems that when this happens, NONMEM uses the default
value (0) for OBJI for the data sent to that process, resulting in a lower OBJ.
So, you can, in principle, rerun the model using the MSF and single process.
But, interestingly, while the OBJ reported in the outfile files seems to be
wrong, NONMEM will still report SEE etc. Can the SEE values in this case be
believed?
Mark Sale M.D.
Senior Vice President, Pharmacometrics
Nuventra Pharma Sciences, Inc.
2525 Meridian Parkway, Suite 200
Durham, NC 27713
Phone (919)-973-0383
[email protected]