NONMEM Users Network Archive

Hosted by Cognigen

RE: OMEGA priors using modes of inverse Wishart matrix

From: Joachim Grevel <jgrevel>
Date: Fri, 24 Feb 2012 12:37:06 -0000

Dear all,

 

The NONMEM control file of Example 1 in the USERs GUIDE to NONMEM 7.2 =
pp.
108/128 tells us how to work with priors:

 

; Prior information to the OMEGAS (NETPxNETP of them).

$OMEGA BLOCK(4)

0.2 FIX

0.0 0.2

0.0 0.0 0.2

0.0 0.0 0.0 0.2

; Degrees of freedom to prior OMEGA matrix (1 for each Omega Prior =
block).

; Because degrees of freedom is very low, equal to the

; the dimension of the prior OMEGA, this means that the prior =
information to
the OMEGAS is

; highly uninformative

$THETA (4 FIX)

 

This is what I use with NONMEM 7.2

 

; when I follow the instruction and use inverse Wishart to represent the
OMEGA of my previous NONMEM analysis (2500 conc in 200 patients).

$OMEGA BLOCK(2)

0.000135 FIX

0.0000195 0.000640

; I want to make the priors informative and calculated df according to =
the
proposal by Mats.

$THETA (138 FIX)

 

; when I do not follow the instruction and use simply the covariance =
matrix
of the previous NONMEM analysis of the large data file:

$OMEGA BLOCK(2)

0.000131 FIX

0.0000191 0.000627

$THETA (138 FIX)

 

There is not really a big difference in my example.

 

Joachim

 

Joachim Grevel, PhD

Scientific Director

BAST Inc Limited

BioCity Nottingham

Pennyfoot Street

Nottingham, NG1 1GF

Tel: +44 (0)115 8120497

 

 

 

 

From: owner-nmusers
On
Behalf Of Stephen Duffull
Sent: 22 February 2012 17:46
To: Mats Karlsson; 'Joachim Grevel'; 'Coen van Hasselt'; 'nmusers'
Subject: RE: [NMusers] OMEGA priors using modes of inverse Wishart =
matrix

 

Hi

 

An appropriate value for dof of the IW is difficult to determine. While =
it
can be set at n-1 from a prior this is somewhat arbitrary. It is not in
this sense like a t-distn where we calculate dof in this manner.

 

You would have to get a feeling for the degree of spread in your =
deviates
given your guess at OMEGA(0) and dof. This can be done by simulation or =
in
special circumstances by direct calculation.

 

Steve

--

 

 

From: owner-nmusers
On
Behalf Of Mats Karlsson
Sent: Thursday, 23 February 2012 3:58 a.m.
To: 'Joachim Grevel'; 'Coen van Hasselt'; 'nmusers'
Subject: RE: [NMusers] OMEGA priors using modes of inverse Wishart =
matrix

 

Dear Joachim,

 

The IW distribution is not something you get from NONMEM, you have to =
give
the degrees of freedom of the IW prior distribution. Normally this is at =
or
below the number of subjects in you previous study depending the =
information
about the parameter per subject.

 

In addition to Userís Guides, you may find useful info in Gisleskog et =
al J
Pharmacokinet Pharmacodyn. 2002 Dec;29(5-6):473-505.

 

Best regards,

Mats

 

Mats Karlsson, PhD

Professor of Pharmacometrics

Dept of Pharmaceutical Biosciences

Faculty of Pharmacy

Uppsala University

Box 591

75124 Uppsala

 

Phone: +46 18 4714105

Fax + 46 18 4714003

 

From: owner-nmusers
On
Behalf Of Joachim Grevel
Sent: 22 February 2012 15:24
To: 'Coen van Hasselt'; nmusers
Subject: RE: [NMusers] OMEGA priors using modes of inverse Wishart =
matrix

 

Dear Coen,

 

There I do have the answer: with MCMC Bayesian in NONMEM 7.1 you have to =
use
NWPRI according to the guide.

 

Dear Nidal,

 

What I try to do is that: use an existing very well defined popPK model
(2500 conc in 200 patients) to obtain individual PK parameters in only =
20
additional patients that have sparse sampling (2 to 4 conc per patient). =
I
was planning to use informative priors rather than add 80 conc to a =
bulk of
2500 conc. What do you think?

 

Thanks to all, specifically Tim!

 

Joachim

 

From: Coen van Hasselt [mailto:Coen.vanHasselt
Sent: 22 February 2012 14:03
To: jgrevel
Subject: Re: [NMusers] OMEGA priors using modes of inverse Wishart =
matrix

 

Dear Joachim,

I have always wondered about this particular question myself as well..
Thanks for asking at NMusers.

Another thing related to the PRIORs I was wondering about: you can =
either
use NWPRI (i.e. with the inverse wishart for OMEGA's), or TNPRI, which =
uses
a multivariate normal for the OMEGA's. Do you have any idea when to use
either of these two possible implementations ?

Thanks,
Coen

>>> "Joachim Grevel" 02/22/12 2:11 PM >>>

Dear NMusers,

Our NONMEM User Guides are full of good advice and they are searchable. =
Yet
when I tried to find out whether any of the estimation methods could be
enticed to give me the ?inverse Wishart matrix? needed to specify PRIORS =
for
OMEGAs, I found no help. Leonid G. mentions the inverse Wishart matrix =
in
his historic contributions to NMusers, but again I could not find out =
how to
obtain it. If it does not happen in NONMEM, is there a tool in R that =
can
give me that matrix?

Your help is greatly appreciated,

 

Joachim

 

Joachim Grevel, PhD

Scientific Director

BAST Inc Limited

BioCity Nottingham

Pennyfoot Street

Nottingham, NG1 1GF

Tel: +44 (0)115 8120497

 



Received on Fri Feb 24 2012 - 07:37:06 EST

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.