Dear all,
we are testing our installation of NONMEM 7.1.2. However we have problems
when we execute the provided examples 10 and 10l from the examples
directory.
Both examples terminate for the last estimation method (Laplacian
Conditional Estimation with Interaction) with a POSTERIOR DENSITY IS
NON-POSITIVE-DEFINITE message during the initial iterations. Looking at
the reference example results we find that in example10.res and
example10l.res there are gradients that are 0 from the beginning and some
messages from NMTRAN (warnings 31 + 41) which indicate the presence of
unused parameters in the model (??) Looking at the provided ctl files we
found that the prior information for OMEGA is missing the FIX attribute.
However when we correct the ctl files accordingly, example 10 still
terminates with the above message, while example 10l now runs fine (even
able to complete the covariance step).
As anyone else also come across this issue? How serious would you evaluate
the fact that our installation of NONMEM behaves differently with regard
to installation qualification?
Thank you in advance and best regards
Matthias Frede
_________________________________________
Bayer Schering Pharma Aktiengesellschaft
BSP-GDD-CP-TS-M&S-M&S CV/H
Phone: +49 202 36 4830
Fax: +49 202 36 3998
E-mail: [email protected]
Web: http://www.bayerscheringpharma.de
Vorstand: Andreas Fibig, Vorsitzender | Andreas Busch, Ulrich Köstlin,
Kemal Malik, Bernd Metzner
Vorsitzender des Aufsichtsrats: Richard Pott
Sitz der Gesellschaft: Berlin | Eintragung: Amtsgericht Charlottenburg HRB
283 B
NONMEM 7: Problems with example10.ctl and example10l.ctl
Dear Dr. Frede:
I too obtain an error from the Laplace method when correcting example10.ctl
with a FIX option on the prior to $OMEGA. This does not reflect an
installation issue, only an issue with the Laplace method, in that it
occasionally is very sensitive to starting positions. In our next version,
7.2.0, we are adding an option to make the Laplace method a little more robust
regarding positive definite Hessian signals that occur in the first iteration.
Meanwhile, example10l, which is identical to example10 but better
parameterized, is a suitable control stream test.
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 [email protected]
Sent: Wednesday, February 02, 2011 11:59 AM
To: [email protected]
Subject: [NMusers] NONMEM 7: Problems with example10.ctl and example10l.ctl
Dear all,
we are testing our installation of NONMEM 7.1.2. However we have problems when
we execute the provided examples 10 and 10l from the examples directory.
Both examples terminate for the last estimation method (Laplacian Conditional
Estimation with Interaction) with a POSTERIOR DENSITY IS NON-POSITIVE-DEFINITE
message during the initial iterations. Looking at the reference example results
we find that in example10.res and example10l.res there are gradients that are 0
from the beginning and some messages from NMTRAN (warnings 31 + 41) which
indicate the presence of unused parameters in the model (??) Looking at the
provided ctl files we found that the prior information for OMEGA is missing the
FIX attribute.
However when we correct the ctl files accordingly, example 10 still terminates
with the above message, while example 10l now runs fine (even able to complete
the covariance step).
As anyone else also come across this issue? How serious would you evaluate the
fact that our installation of NONMEM behaves differently with regard to
installation qualification?
Thank you in advance and best regards
Matthias Frede
_________________________________________
Bayer Schering Pharma Aktiengesellschaft
BSP-GDD-CP-TS-M&S-M&S CV/H
Phone: +49 202 36 4830
Fax: +49 202 36 3998
E-mail: [email protected]
Web: http://www.bayerscheringpharma.de
Vorstand: Andreas Fibig, Vorsitzender | Andreas Busch, Ulrich Köstlin, Kemal
Malik, Bernd Metzner
Vorsitzender des Aufsichtsrats: Richard Pott
Sitz der Gesellschaft: Berlin | Eintragung: Amtsgericht Charlottenburg HRB 283 B