NONMEM Users Network Archive

Hosted by Cognigen

RE: NONMEM vs SPSS

From: E.Olofsen
Date: Mon, 31 Mar 2014 12:23:06 +0000

Hi Jeroen,

I've just tried this with the CONTROL3 example, and with a $TABLE record. W=
hile there are randomly sampled observation records in the output of $TABLE=
, there are also zero, one, or more than one dose records, and estimation s=
teps with zero gradients if there is no dose record. So it seems that this =
approach cannot be used at present.

Erik
________________________________
From: Elassaiss - Schaap, J [jeroen.elassaiss-schaap
Sent: Monday, March 31, 2014 1:29 PM
To: Olofsen, E. (ANST); gej1000
Subject: RE: [NMusers] NONMEM vs SPSS

Hi Erik,

Thanks! I never explored the BOOTSTRAP option in NM7.3. Interesting.

Thinking of it, not sure how to interpret such bootstrap results unless the=
re is plenty of sampling from that individual relative to model complexity.=
 Perhaps a classical cross-validation would be my preferred option, but one=
 could also consider to bootstrap residuals.

Jeroen


From: E.Olofsen
Sent: Monday, March 31, 2014 13:08
To: Elassaiss - Schaap, J (Jeroen); gej1000
om
Subject: RE: [NMusers] NONMEM vs SPSS

Hi Jeroen,

The BOOTSTRAP option of $SIMULATION gives different results when N=1 (eac=
h measurement is treated as having a different ID). Could that perhaps be u=
seful?

Erik
________________________________
From: owner-nmusers
ner-nmusers
roen.elassaiss-schaap
Sent: Monday, March 31, 2014 12:23 PM
To: Gavin Jarvis; nmusers
Subject: RE: [NMusers] NONMEM vs SPSS
Dear Gavin,

Reading back your original post, if your data are really N=1 and you have=
 this perfect fit phenomenon there is probably little value in reporting th=
e SEs. But on the other hand your new reply suggests that you are doing a s=
imulation exercise… in which case a regression on aggregated data may be =
less productive. Perhaps you could consider doing an analysis with SSE (psn=
.sf.net, not sure whether WfN has similar tools) to figure out which design=
 would support the models you are considering with little additional effort=
.

Jeroen

PS: bootstrap on N=1 does not work, the nonmem approaches use all samplin=
g over subjects. (There are other ways of doing a bootstrap)

From: owner-nmusers
ilto:owner-nmusers
Sent: Monday, March 31, 2014 11:33
To: nmusers
Subject: RE: [NMusers] NONMEM vs SPSS

Dear All

Thank you for all the very helpful comments.

In reply:

1. MATRIX=R does make the standard error and correlation values muc=
h more similar to SPSS(NLR)

2. The residual error model is additive, homoscedastic (just ETA(1)).=
 The data are extremely tight (R^2 >99.9%) – almost perfect! The purpose =
of my analysis is to assess structural models for analysing asymmetric dose=
-response curves. The problem is that some models produces parameters that =
lose empirical meaning and are very highly correlated.

3. I tried the bootstrap option using WFN. However, the parameter est=
imates all came out identical – probably because the data is so tight –=
 this makes it tricky to evaluate standard errors!

Gavin


From: Bauer, Robert [mailto:Robert.Bauer
Sent: 29 March 2014 20:46
To: Ken Kowalski; 'Gavin Jarvis'; nmusers
bomaxnm.com>
Subject: RE: [NMusers] NONMEM vs SPSS

I concur with Ken’s statement, and I also prefer to use MATRIX=R as the=
 first choice for covariance assessment. On occasion, MATRIX=S can be us=
ed if there are numerical difficulties in assessing the R matrix, and if th=
ere are enough subjects relative to the dimension size (number of total par=
ameters estimated) of the variance-covariance matrix to be estimated.

Robert J. Bauer, Ph.D.
Vice President, Pharmacometrics, R&D
ICON Development Solutions
7740 Milestone Parkway
Suite 150
Hanover, MD 21076
Tel: (215) 616-6428
Mob: (925) 286-0769
Email: Robert.Bauer
Web: www.iconplc.com<http://www.iconplc.com/>

From: owner-nmusers
ilto:owner-nmusers
Sent: Saturday, March 29, 2014 3:44 PM
To: 'Gavin Jarvis'; nmusers
Subject: RE: [NMusers] NONMEM vs SPSS

Dear Gavin,

This is most likely because most nonlinear regression programs invert the H=
essian (second derivative matrix of the model with respect to the parameter=
s) to obtain the covariance matrix. This corresponds to the R matrix in NO=
NMEM. However, the default method that NONMEM uses is a sandwich estimator=
 involving both the Hessian (R) and the square of the first derivatives mat=
rix (S). I suspect that if you use the MATRIX=R option on the $COV step =
you will find that the standard errors will now be in agreement with SPSS (=
NLR). I know Stu Beal made the sandwich estimator the default as it is sup=
posed to be more robust to non-normality but I would have preferred the MAT=
RIX=R option to be the default to be more consistent with other nonlinear=
 regression software implementations.
Ken

From: owner-nmusers
ilto:owner-nmusers
Sent: Saturday, March 29, 2014 12:55 PM
To: nmusers
Subject: [NMusers] NONMEM vs SPSS

Dear NONMEM Users

Does anyone have a view on the relative merits/reliability/accuracy of NONM=
EM ($COV step) vs SPSS (NLR) with respect to their derived values of the pa=
rameter standard errors and parameter correlation matrices?

The data I am analysing are single subject (not population). Parameter esti=
mates from the two programs are, to all intents and purposes, identical. Ho=
wever, the SE values from NONMEM $COV are consistently smaller by 1.5-2.0-f=
old.

Any thoughts?

Gavin


__________________________________________________
Dr Gavin E Jarvis MA PhD VetMB MRCVS
University Lecturer in Veterinary Anatomy
Department of Physiology, Development & Neuroscience
Physiological Laboratory
Downing Street
Cambridge
CB2 3EG
Tel: +44 (0) 1223 333745

Fellow and College Lecturer in Pharmacology
Selwyn College
Cambridge
CB3 9DQ
Tel: +44 (0) 1223 761303

Email: gej1000
Web: www.pdn.cam.ac.uk/staff/jarvis<http://www.pdn.cam.ac.uk/staff/jarvis>
Twit:



ICON plc made the following annotations.
---------------------------------------------------------------------------=
---
This e-mail transmission may contain confidential or legally privileged inf=
ormation that is intended only for the individual or entity named in the e-=
mail address. If you are not the intended recipient, you are hereby notifie=
d that any disclosure, copying, distribution, or reliance upon the contents=
 of this e-mail is strictly prohibited. If you have received this e-mail tr=
ansmission in error, please reply to the sender, so that ICON plc can arran=
ge for proper delivery, and then please delete the message.

Thank You,

ICON plc
South County Business Park
Leopardstown
Dublin 18
Ireland
Registered number: 145835

Notice: This e-mail message, together with any attachments, contains
information of Merck & Co., Inc. (One Merck Drive, Whitehouse Station,
New Jersey, USA 08889), and/or its affiliates Direct contact information
for affiliates is available at
http://www.merck.com/contact/contacts.html) that may be confidential,
proprietary copyrighted and/or legally privileged. It is intended solely
for the use of the individual or entity named on this message. If you are
not the intended recipient, and have received this message in error,
please notify us immediately by reply e-mail and then delete it from
your system.

Notice: This e-mail message, together with any attachments, contains
information of Merck & Co., Inc. (One Merck Drive, Whitehouse Station,
New Jersey, USA 08889), and/or its affiliates Direct contact information
for affiliates is available at
http://www.merck.com/contact/contacts.html) that may be confidential,
proprietary copyrighted and/or legally privileged. It is intended solely
for the use of the individual or entity named on this message. If you are
not the intended recipient, and have received this message in error,
please notify us immediately by reply e-mail and then delete it from
your system.

Received on Mon Mar 31 2014 - 08:23:06 EDT

The NONMEM Users Network is maintained by ICON plc. Requests to subscribe to the network should be sent to: nmusers-request@iconplc.com.

Once subscribed, you may contribute to the discussion by emailing: nmusers@globomaxnm.com.