From: "Hu, Chuanpu"
Subject: [NMusers] Simultaneous modeling of PK and categorical PD
Date: Tue, 2 Apr 2002 10:24:32 -0500
Dear All,
I am trying to model PK and a categorical PD endpoint simultaneously, and
thought I could achieve that by specifying the likelihood for PK and PD.
However I do not see how to make NONMEM accept the normal likelihood. Just
for a start, I simulated a simple 1-compartment population data, and tried
to use ADVAN1 with
$ERROR
Y = 1/THETA(.) * EXP( -( (DV-F)/THETA(.) )**2/2 )
...
$EST MET=1 LAPLACE LIKE MAX=500
and got an error message that I have not seen before:
ERROR IN CLIK WITH INDIVIDUAL 1 ID=0.10000000E+01
CONDITIONAL LIKELIHOOD FOR SOME OBSERVATION ESTIMATED TO BE NEGLIGIBLE
I tried to replace the estimation step with simulation, but NONMEM did not
like it either (I am guessing because there is no EPS(). ).
Can anyone tell me how to specify the continuous portion of the likelihood?
(I plan to specify the PD likelihood according to the logistic regression
example.)
Thanks,
Chuanpu
----------------------------------------------------------------
Chuanpu Hu, Ph.D.
Modeling and Simulation, U.S.
CPDM
GlaxoSmithKline
Tel: 919-483-8205
Fax: 919-483-6380
Simultaneous modeling of PK and categorical PD
From: Nick Holford
Subject: RE: [NMusers] Simultaneous modeling of PK and categorical PD
Date: Wed, 03 Apr 2002 05:16:07 +1200
Chuanpu,
Stuart Beal revealed the trick for simultaneous models of continuous and categorical data to Tom
Ludden a couple of year ago and Tom was kind enough to tell me. I would like to put the details on
the NONMEM repository but cannot find any links to the repository that are working. If someone can
tell me where to put the file and make a link to the repository work e.g. from the Globomax home
page then you can see the solution to this problem.
Nick
--
Nick Holford, Divn Pharmacology & Clinical Pharmacology
University of Auckland, 85 Park Rd, Private Bag 92019, Auckland, New Zealand
email:n.holford@auckland.ac.nz tel:+64(9)373-7599x6730 fax:373-7556
http://www.phm.auckland.ac.nz/Staff/NHolford/nholford.htm
From: Nelamangala V Nagaraja
Subject: RE: [NMusers] Simultaneous modeling of PK and categorical PD
Date: Tue, 02 Apr 2002 12:40:53 -0500
There is one link at
http://anesthesia.stanford.edu/pkpd/NONMEM%20Repository/
Raj
From: Nick Holford
Subject: Re: [NMusers] Simultaneous modeling of PK and categorical PD
Date: Wed, 03 Apr 2002 05:57:43 +1200
Thanks but this link gives no instructions for how to upload files....
--
Nick Holford, Divn Pharmacology & Clinical Pharmacology
University of Auckland, 85 Park Rd, Private Bag 92019, Auckland, New Zealand
email:n.holford@auckland.ac.nz tel:+64(9)373-7599x6730 fax:373-7556
http://www.phm.auckland.ac.nz/Staff/NHolford/nholford.htm
From: LSheiner
Subject: [NMusers] Likelihood
Date: Tue, 02 Apr 2002 13:44:22 -0800
FYI ...
Quoted reply history
-------- Original Message --------
Date: Tue, 2 Apr 2002 09:26:32 -0800 (PST)
From: "S.Beal"
To: lewis
Assuming that there is a covariate being tested that allows Y to be set
as above for a PK observation and differently for a PD observation,
in principle the above should work. There could be any number of
reasons peculiar to the user's set-up that accounts for the error message.
It is possible that the argument of EXP is becoming very negative, and
due to the heavy-truncation of the EXP function, the likelihood has
become 0. It might be better to use the option -2LL, rather than LIKE
(and code the above accordingly).
Stu
From: "Hu, Chuanpu"
Subject: RE: [NMusers] Likelihood
Date: Tue, 2 Apr 2002 17:47:14 -0500
Using -2LL works. Thanks a lot to Lew, Stu and Shibing Deng. I should have
guessed this from the error message.
Chuanpu
From: Nick Holford
Subject: [NMusers] Simultaneous Continuous and Categorical Modelling
Date: Thu, 04 Apr 2002 00:29:22 +1200
Hi,
Bill Bachman has promptly responded to my reqest to add files illustrating a general method for
doing Simultaneous Continuous and Categorical Modelling. Thanks Bill.
There are two files, readme.txt which gives an overview of the method and how it was validated by
simulation, and contodd.zip which contains control streams and data for the example as well as
other material showing how the validation was done.
Please send comments on how to improve the readme.txt file to make it clearer and more helpful to
me.
Nick
"Bachman, William" wrote:
>
> Nick,
>
> Files have been placed in:
ftp://ftp.globomaxnm.com/Public/nonmem/continuous&categorical/
>
> please check access
>
> Bill
>
--
Nick Holford, Divn Pharmacology & Clinical Pharmacology
University of Auckland, 85 Park Rd, Private Bag 92019, Auckland, New Zealand
email:n.holford@auckland.ac.nz tel:+64(9)373-7599x6730 fax:373-7556
http://www.phm.auckland.ac.nz/Staff/NHolford/nholford.htm
********