NONMEM Users Network Archive

Hosted by Cognigen

RE: No TABLE output with 1500 measurements per individual

From: Jean-Marie.Martinez
Date: Thu, 24 Nov 2016 16:45:17 +0000

Dear Pieter,

Your trick worked fine : the duration of the table creation was reduced fro=
m (theoretical) 70 days to one hour !

Thank you very much :)

Kind regards,

Jean-Marie Martinez
Modeling & Simulation Group
Sanofi Montpellier

De : Pieter Colin [mailto:Pieter.Colin
Envoyé : mercredi 23 novembre 2016 11:19
À : Martinez, Jean-Marie R&D/FR; nmusers
Objet : RE: No TABLE output with 1500 measurements per individual

Dear Jean-Marie Martinez,

We've encountered a similar problem in the past.
Per my understanding calculation of the CWRES might be the problem here.
By adding WRESCHOL as option in your first table, after the "FILE="-state=
ment, this problem should be resolved.

Kind regards,

Pieter Colin
University Medical Center Groningen
Department of Anesthesiology (EB32)

From: owner-nmusers
Sent: woensdag 23 november 2016 9:16
To: nmusers
Subject: [NMusers] RE: No TABLE output with 1500 measurements per individua=

Dear NM-Users,

We are trying to model with NONMEM 7.3 (FOCE.I) a dataset containing one me=
asurement per minute for 68 individuals, for a total of 1440 measurements p=
er individual. The dataset is therefore composed of ~100,000 rows.
After having defined $SIZES adequately (LIM6 statement), the (successful) m=
inimization of the algorithm takes few minutes, without any warning or erro=
r messages.

The problem is that the TABLE output step is nearly impossible to obtain. E=
ven with the FIRSTONLY statement, no table can be obtained in a reasonable =
time period.

Trying to limit the table contents to ID & IPRED only does not solve the is=

We performed some additional tests with MSF files as input. When applying M=
SF to a dataset composed of only ONE individual at a time, the table takes =
24 hours to be generated.
This would therefore take 68 days (if time increases proportionally... ?) t=
o obtain PRED & IPRED for all individuals.

When implementing the model in an alternative commercial software, the tabl=
e is output in less than 20 minutes.

A solution would be to use MSF on a reduced dataset i.e., with a (randomly)=
 decreased number of measurements per individual, but we want to avoid this=
What other solution do we have ?

Can anyone provide some input on it ?

Thanks !

Jean-Marie Martinez
Modeling & Simulation Group
Sanofi Montpellier

Received on Thu Nov 24 2016 - 11:45:17 EST

The NONMEM Users Network is maintained by ICON plc. Requests to subscribe to the network should be sent to:

Once subscribed, you may contribute to the discussion by emailing: