RE: RANMETHOD also for $SIM
Bob and Sven,
It's not a question of 'mathematical correctness' , but to what use you intend
to put the simulated data. Quasi-random sequences are much different than
pseudo-random sequences and any statistical conclusions you might want to draw
from the simulated data may be dubious. For some purposes such as numerical
integration, quasi-random etas can greatly out perform
pseudo-random etas. But for other purposes, the lack of 'randomness' in
quasi-random sequences can make them quite inappropriate. For example,
It is well known that raw (unscrambled) quasi -random sequences cannot be
substituted for pseudo-random sequences in MCMC applications such as SAEM -
results can be
quite nonsensical. Bob's claim that ""It (use of SOBOL sequnces) would even
be inappropriate (that is, lead to bias) in the SAEM or BAYES method"
is certainly correct for raw SOBOL sequences.
However, scrambling (for example, Owen scrambling given by the S1 option in NM
) may overcome this problem in the MCMC case and currently this is
a research topic of great interest in the MCMC world. For example, Owen and
Tribble
(PNAS , Vol 102, 2005, 8844-8849,
http://www.pnas.org/content/102/25/8844.long) have recently proved that under
some conditions a Metropolis-Hastings algorithm
with quasi-random inputs yields consistent estimates and the method is much
more accurate than ordinary Metroplis-Hastings sampling with standard
pseudo-random inputs.
They also show how to extend the results to Gibbs sampling.
Quoted reply history
From: [email protected] [mailto:[email protected]] On
Behalf Of Bauer, Robert
Sent: Wednesday, October 10, 2012 11:34 AM
To: Mensing, Sven; [email protected]
Subject: [NMusers] RE: RANMETHOD also for $SIM
Sven:
It would be mathematically incorrect to create simulated values using a Sobol
quasi-random method. You would obtain eta and eps values that are not truly
normally distributed. The Sobol technique is specifically designed to improve
the efficiency of Monte Carlo integration in multi-dimensional space, such as
is done during the expectation step of the Monte Carlo importance sampling EM
methods, and for no other purpose. It would even be inappropriate (that is,
lead to bias) in the SAEM or BAYES methods. This is why the Sobol technique is
only for IMP and DIRECT estimation.
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: [email protected]<mailto:[email protected]>
Web: http://www.iconplc.com/
________________________________
From: [email protected]<mailto:[email protected]>
[mailto:[email protected]]<mailto:[mailto:[email protected]]>
On Behalf Of Mensing, Sven
Sent: Wednesday, October 10, 2012 8:57 AM
To: [email protected]<mailto:[email protected]>
Subject: [NMusers] RANMETHOD also for $SIM
Hello,
is it possible to use the Sobol pseudo-random number generator (e.g.
RANMETHOD=2S) for $SIMULATION problems? Maybe in NM 7.3?
Thanks
Sven
________________________________
Abbott GmbH & Co. KG
Sitz der Gesellschaft: Wiesbaden - Registergericht: AG Wiesbaden HRA 4888
Persönlich haftende Gesellschafterin: Abbott Management GmbH
Sitz der persönlich haftenden Gesellschafterin: Wiesbaden - Registergericht: AG
Wiesbaden HRB 12889
Geschäftsführer: Alexander Würfel, Dr. Friedrich Richter, Matthias Däschner ,
Kathy Vinchkoski Turner
Vorsitzender des Aufsichtrats: Brian Blaser
This communication may contain information that is proprietary, confidential,
or exempt from disclosure. If you are not the intended recipient, please note
that any other dissemination, distribution, use or copying of this
communication is strictly prohibited. Anyone who receives this message in error
should notify the sender immediately by telephone or by return e-mail and
delete it from his or her computer.
Diese Kommunikation kann Informationen enthalten, die geheim, vertraulich oder
hinsichtlich der Offenlegung beschränkt sind. Wenn Sie nicht der beabsichtigte
Empfänger sind, nehmen Sie bitte zur Kenntnis, dass jede Weitergabe,
Verteilung, Verwendung oder Vervielfältigung dieser. Kommunikation strikt
untersagt ist. Jeder, der diese Nachricht fehlerhaft erhält, sollte den Sender
unverzüglich telefonisch oder durch Rücksendung der E-Mail benachrichtigen und
diese von seinem oder ihrem Computer löschen.