From: "Pravin Jadhav" pravinj@gmail.com
Subject: [NMusers] NONMEM troubleshoot
Date: Tue, October 19, 2004 11:11 pm
Hello all,
I am getting some strange behavior while running one problem set. I am
performing sequential PK-PD modeling for a drug I am working on. The
data consist of 99 patients studied over 3 periods with a total 36000
PD observations.
Recently, I had to recompile NONMEM to run this problem. I ran a base
model and got results closer to my expectations. After updating the
model, I run into this strange problem:- NONMEM seems to run thr' its
iterations (as I see on my screen), but at one point gets stuck
FOREVER.
It looks like it was closer to the minimum (guess based on a little
change in the OFV among iterations, which was also less than the OFV
for the base model). It may be a local minimum but I tried
constraining the parameters within realistic range w/o success. I know
sounds very vague but I have no better explanation. I just want to
know where to look for with such kind of troubleshooting. BTW, I am
running METH=0 and not performing $COV step.
Thanks,
Pravin
--
Pravin Jadhav
Graduate Student
Department of Pharmaceutics
MCV/Virginia Commonwealth University
DPE1/CDER/OCPB/Food and Drug Administration
Phone: (301) 594-2623
Fax: (301) 480-3212
NONMEM troubleshoot
From: Chuanpu.Hu@sanofi-synthelabo.com
Subject: RE: [NMusers] NONMEM troubleshoot
Date: Thu, October 21, 2004 9:25 am
Dear Pravin,
I have had NONMEM stuck with "normal" PK data, as others did too. Whether
it is near a local minimum or not does not seem to matter. It happens very
rarely and changing initial estimates always solved the problem. It sounds
like your problem is more difficult, but just wondering whether you have
tried it. By the way, do the initial estimates corresponding to reasonable
goodness of fit?
Good luck,
Chuanpu
-------------------------------------------------------------------
Chuanpu Hu, Ph.D.
Assistant Director, Biostatistics
Sanofi-Synthelabo Research
9 Great Valley Parkway
Malvern, PA 19355-1304
Tel: (610) 889-6774
Fax: (610) 889-6932
From:"Ruedi Port" r.port@dkfz.de
Subject: RE: [NMusers] NONMEM troubleshoot
Date: Fri, October 22, 2004 3:59 am
Hi Pravin:
As Chuanpu says, NONMEM getting stuck without telling why
is not uncommon. You can see where NONMEM was just before
it got silent, in the following ways:
1) During iterations, the current parameter estimates are
continuously written to the file "INTER". That file might
show you some crazy estimate before NONMEM gets stuck.
2) You could set $ESTIMATION MAXEVALS to a number such
that it stops just after the last successful iteration,
then graphically inspect the predictions.
One or two wrong measurements that are far from reality
can cause lots of trouble by inflating the estimate of a
variance parameter.
Good luck!
Ruedi
From: "Pravin Jadhav" pravinj@gmail.com
Subject: RE: [NMusers] NONMEM troubleshoot
Date: Fri, October 22, 2004 1:07 pm
Thanks to Ruedi, Chuanpu and Nitin for their replies.
Actually, the mistake was so silly that I should have managed it in
the first place. I was just playing around with different things ( I
tried different initial estimates, boundaries etc. etc.) and this is
what I find! I had 23 variables in $TABLE block and I did not specify
NOAPPEND(I forgot). Upon writing NOAPPEND everything looks fine and I
have my output. Basically, NONMEM was getting stuck at $TABLE step (I
think).
******What I don't understand is why didn't I get an error
message?***** However, I do remember getting such error messages in
the past for equivalent syntax
problem. I don't know what happened this time. The only difference
this time is I have a large dataset and a slightly complex PD model.
Upon Nick's advice, I have reported it to nmconsult@globomaxnm.com.
Seems like Nick also had similar trouble in the past.
Thank you.
Pravin
From: "Bachman, William (MYD)" bachmanw@iconus.com
Subject: RE: [NMusers] NONMEM troubleshoot
Date: Fri, October 22, 2004 1:25 pm
When reporting such "anomolies", you need to supply a control stream and
sufficient data to reproduce the issue. Otherwise, it's a wild goose chase
for nmconsult. Only sufficient data to generate the issue is necessary. If
concerned about proprietary data, be sure to mask your data.
nmconsult@globomaxnm.com
GloboMax
The Strategic Pharmaceutical Development Division of ICON plc
7250 Parkway Drive, Suite 430
Hanover, MD 21076
Voice: (410) 782-2205
FAX: (410) 712-0737
_______________________________________________________