Dear uses,
I am using NM7 SAEM method for an analysis. As I learned from NM7 workshop, I
used IMP method after SAEM to generate SE estimate.
In this case, at the end of SAEM iteration, OBJ was quite stable and negative
several thousand.
But, with IMP, after 5 iterations, Obj was up to positive several 10 to power
of 7. Obj was decreasing during this 5 iterations.
My questions is, dose this mean SE estimate is not reliable? Should I
increase
iterations of IMP?
Thank you.
quality of SE estimate by IMP method
At this point the SE may be unreliable. Try using IMPMAP (and EONLY=1)
and see if that helps. Let it run for 20 iterations.
Robert J. Bauer, Ph.D.
Vice President, Pharmacometrics
ICON Development Solutions
Tel: (215) 616-6428
Mob: (925) 286-0769
Email: [email protected]
Web: www.icondevsolutions.com
Quoted reply history
________________________________
From: [email protected] [mailto:[email protected]]
On Behalf Of Ethan Wu
Sent: Thursday, September 16, 2010 11:46 AM
To: [email protected]
Subject: [NMusers] quality of SE estimate by IMP method
Dear uses,
I am using NM7 SAEM method for an analysis. As I learned from NM7
workshop, I used IMP method after SAEM to generate SE estimate.
In this case, at the end of SAEM iteration, OBJ was quite stable and
negative several thousand.
But, with IMP, after 5 iterations, Obj was up to positive several 10
to power of 7. Obj was decreasing during this 5 iterations.
My questions is, dose this mean SE estimate is not reliable? Should I
increase iterations of IMP?
Thank you.
Hi Bob,
IMP/MAP also came back with pretty similar OBJ like IMP.
Below is the code I use, am I doing the wrong way?
$ESTIMATION METHOD=IMPMAP INTERACTION LAP EONLY=1 NOABORT NITER=20
PRINT=1 SIGL=6 NSIG=2 ISAMPLE=1000
Thank you.
Quoted reply history
________________________________
From: "Bauer, Robert" <[email protected]>
To: Ethan Wu <[email protected]>; [email protected]
Sent: Thu, September 16, 2010 2:26:07 PM
Subject: RE: [NMusers] quality of SE estimate by IMP method
Yes. Read the METHOD=CHAIN section in the intro712.pdf manual.
Robert J. Bauer, Ph.D.
Vice President, Pharmacometrics
ICON Development Solutions
Tel:(215) 616-6428
Mob: (925) 286-0769
Email: [email protected]
Web: www.icondevsolutions.com
________________________________
From: Ethan Wu [mailto:[email protected]]
Sent: Thursday, September 16, 2010 12:19 PM
To: Bauer, Robert; [email protected]
Subject: Re: [NMusers] quality of SE estimate by IMP method
Hi Bob,
Is there a way to pick up from the last run with SAEM/IMP so to only run the
additional step IMPMAP?
Thank you.
________________________________
From: "Bauer, Robert" <[email protected]>
To: Ethan Wu <[email protected]>; [email protected]
Sent: Thu, September 16, 2010 12:12:28 PM
Subject: RE: [NMusers] quality of SE estimate by IMP method
At this point the SE may be unreliable. Try using IMPMAP (and EONLY=1) and see
if that helps. Let it run for 20 iterations.
Robert J. Bauer, Ph.D.
Vice President, Pharmacometrics
ICON Development Solutions
Tel:(215) 616-6428
Mob: (925) 286-0769
Email: [email protected]
Web: www.icondevsolutions.com
________________________________
From: [email protected] [mailto:[email protected]] On
Behalf Of Ethan Wu
Sent: Thursday, September 16, 2010 11:46 AM
To: [email protected]
Subject: [NMusers] quality of SE estimate by IMP method
Dear uses,
I am using NM7 SAEM method for an analysis. As I learned from NM7 workshop, I
used IMP method after SAEM to generate SE estimate.
In this case, at the end of SAEM iteration, OBJ was quite stable and negative
several thousand.
But, with IMP, after 5 iterations, Obj was up to positive several 10 to power
of 7. Obj was decreasing during this 5 iterations.
My questions is, dose this mean SE estimate is not reliable? Should I
increase
iterations of IMP?
Thank you.
Ethan:
I have come across this problem of yours from another user. I could not
resolve that one until I made an improvement in the code, which will
come out in January. I recall from that experience that the SAEM result
was fine, but the objective function could not be well determined
because of a peculiarity of a position of a data point, which resulted
in a poor fit from the MAP estimation, which IMP (On the first
iteration) and IMPMAP (on all iterations) rely on to obtain the initial
proposal density. That you are using LAP also suggests this. If you
wish, you may send me your control stream and data set, and I can try a
couple of extra things to see if it will work in nm712.
Robert J. Bauer, Ph.D.
Vice President, Pharmacometrics
ICON Development Solutions
Tel: (215) 616-6428
Mob: (925) 286-0769
Email: [email protected]
Web: www.icondevsolutions.com
Quoted reply history
________________________________
From: Ethan Wu [mailto:[email protected]]
Sent: Friday, September 17, 2010 9:13 AM
To: Bauer, Robert; [email protected]
Subject: Re: [NMusers] quality of SE estimate by IMP method
Hi Bob,
IMP/MAP also came back with pretty similar OBJ like IMP.
Below is the code I use, am I doing the wrong way?
$ESTIMATION METHOD=IMPMAP INTERACTION LAP EONLY=1 NOABORT NITER=20
PRINT=1 SIGL=6 NSIG=2 ISAMPLE=1000
Thank you.
________________________________
From: "Bauer, Robert" <[email protected]>
To: Ethan Wu <[email protected]>; [email protected]
Sent: Thu, September 16, 2010 2:26:07 PM
Subject: RE: [NMusers] quality of SE estimate by IMP method
Yes. Read the METHOD=CHAIN section in the intro712.pdf manual.
Robert J. Bauer, Ph.D.
Vice President, Pharmacometrics
ICON Development Solutions
Tel: (215) 616-6428
Mob: (925) 286-0769
Email: [email protected]
Web: www.icondevsolutions.com http://www.icondevsolutions.com/
________________________________
From: Ethan Wu [mailto:[email protected]]
Sent: Thursday, September 16, 2010 12:19 PM
To: Bauer, Robert; [email protected]
Subject: Re: [NMusers] quality of SE estimate by IMP method
Hi Bob,
Is there a way to pick up from the last run with SAEM/IMP so to only
run the additional step IMPMAP?
Thank you.
________________________________
From: "Bauer, Robert" <[email protected]>
To: Ethan Wu <[email protected]>; [email protected]
Sent: Thu, September 16, 2010 12:12:28 PM
Subject: RE: [NMusers] quality of SE estimate by IMP method
At this point the SE may be unreliable. Try using IMPMAP (and EONLY=1)
and see if that helps. Let it run for 20 iterations.
Robert J. Bauer, Ph.D.
Vice President, Pharmacometrics
ICON Development Solutions
Tel: (215) 616-6428
Mob: (925) 286-0769
Email: [email protected]
Web: www.icondevsolutions.com http://www.icondevsolutions.com/
________________________________
From: [email protected] [mailto:[email protected]]
On Behalf Of Ethan Wu
Sent: Thursday, September 16, 2010 11:46 AM
To: [email protected]
Subject: [NMusers] quality of SE estimate by IMP method
Dear uses,
I am using NM7 SAEM method for an analysis. As I learned from NM7
workshop, I used IMP method after SAEM to generate SE estimate.
In this case, at the end of SAEM iteration, OBJ was quite stable and
negative several thousand.
But, with IMP, after 5 iterations, Obj was up to positive several 10
to power of 7. Obj was decreasing during this 5 iterations.
My questions is, dose this mean SE estimate is not reliable? Should I
increase iterations of IMP?
Thank you.