Hi all,
When I change the initial values, NM will give different estimates, with
large or small differences. My questions are:
1. Which set of estimates can be mostly trusted and taken as the final
estimates?
2. If the estimates change a lot, does it mean that the model is not
stable and need to be improved?
3. When I have poor data ( observations are too little, or BSV is too
big), sometimes I am not sure if the data or the model causes the unstable
of the estimates. How can I know that the best results are reached and no
more work is needed?
Any advice is much appreciated.
Paul
School of Pharmacy
University of Pittsburgh
716 Salk Hall
3501 Terrace Street
Pittsburgh, PA 15261
Phone: 412-648-8546
E-mail: [email protected]
estimates
Hi Paul,
Sensitivity of model outcome to initial estimate is a well-known phenomenon. If
you search the nmusers archive you will find that it has been mentioned many,
many times. This sensitivity holds for any method in non-linear model
estimation. The general idea is that the more stable the model and(/or) the
better the model describes the data, the less sensitive the outcome is to
initial estimates.
You can treat the outcome of different initial estimates as any other model
change that cannot be judged by likelihood change alone. Ask yourself questions
like: Are the resulting parameters reasonable, are there any issues in the
goodness-of-fit plots, etc etc. In absence of any other important differences I
routinely select the one with lowest objective function value.
I therefore would not discard a model purely because of sensitivity to initial
estimates. On the other hand, if a very robust model is called for you would
need to pay a lot of attention to this sensitivity. It again all depends no the
purpose of the model.
Hope this helps,
Jeroen
Modeling & Simulation Expert
Pharmacokinetics, Pharmacodynamics & Pharmacometrics (P3) - DMPK
MSD
PO Box 20 - AP1112
5340 BH Oss
The Netherlands
[email protected]
T: +31 (0)412 66 9320
M: +31 (0)6 46 101 283
F: +31 (0)412 66 2506
www.msd.com
Quoted reply history
-----Original Message-----
From: [email protected] [mailto:[email protected]] On
Behalf Of [email protected]
Sent: donderdag 10 maart 2011 22:56
To: nm nm
Subject: [NMusers] estimates
Hi all,
When I change the initial values, NM will give different estimates, with
large or small differences. My questions are:
1. Which set of estimates can be mostly trusted and taken as the final
estimates?
2. If the estimates change a lot, does it mean that the model is not
stable and need to be improved?
3. When I have poor data ( observations are too little, or BSV is too
big), sometimes I am not sure if the data or the model causes the unstable
of the estimates. How can I know that the best results are reached and no
more work is needed?
Any advice is much appreciated.
Paul
School of Pharmacy
University of Pittsburgh
716 Salk Hall
3501 Terrace Street
Pittsburgh, PA 15261
Phone: 412-648-8546
E-mail: [email protected]
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.