RE: Nonparametric method
Nick,
We have made a note to include something in the $NONPARAMETRIC help item
about active etas.
You wrote:
>>What is strange here is the error message seems to anticipate what you
are now identifying as a bug. If you fix the bug >>with NP is there a
situation where this error message can occur?
This is one of the issues that is causing us to be cautious about
providing a fix. The NP routine counts the number of individuals with
at least one active eta. The value of this variable is later used as a
denominator. Prior to this use the variable is tested and if it is zero
the "no individuals with active eta's" smessage is issued and the
routine terminates. I do not know if Stuart Beal was just being
cautious and preventing any chance for a divide-by-zero run-time error
or if he had in mind a particular set of circumstances that might result
in this condition. It appears from the code and from tracing variable
values for Rik Schoemaker's example that under some circumstances there
can be a failure to correctly reset the base number of active eta's.
There also appears to be a problem associated with using the INTERACTION
option when individuals do not have the same active eta's.
Alison Boeckmann has offered to look at these problems when time
permits. I have sent her the results obtained with various code changes
for Rik's example, for test problems provided by Mats Karlsson and for
test problems constructed here.
If anyone besides Rik has encountered the "no individuals with active
eta's" message we would like to know the circumstances.
Thanks
Tom
Quoted reply history
-----Original Message-----
From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]
On Behalf Of Nick Holford
Sent: Tuesday, July 24, 2007 4:52 PM
To: [email protected]
Subject: Re: [NMusers] Nonparametric method
Tom,
Rik Schoemaker wrote:
> > I'm trying to implement nonparametric population parameter
estimation in NONMEM VI but I run into the following error message:
> >
> > NONPARAMETRIC ESTIMATE REQUESTED. BUT NO INDIVIDUAL RECORDS
EXIST
> > WITH DATA MODELED WITH ACTIVE ETA'S
You wrote:
> > Since Rik Schoemaker's original e-mail a few weeks ago, the code
> > associated with the NONPARAMETRIC option has been reviewed. There
are
> > problems (bugs) that may reveal themselves when the data set
contains
> > individuals with different active etas. For example, a data set
with
> > some individuals receiving only IV doses while others receive only
oral
> > doses may exhibit this bug. For the individuals receiving only IV
> > doses, the eta's associated with absorption-related parameter such
> > as F1, ALAG1, KA will be nonactive.
This example makes it reasonably clear what an 'ACTIVE ETA' is but there
is no mention of 'ACTIVE ETA' in the NONMEM help. Perhaps you could
consider adding this to the next release so that users can get some
clues on how to interpret this error message.
What is strange here is the error message seems to anticipate what you
are now identifying as a bug. If you fix the bug with NP is there a
situation where this error message can occur?
Nick
--
Nick Holford, Dept Pharmacology & Clinical Pharmacology
University of Auckland, 85 Park Rd, Private Bag 92019, Auckland, New
Zealand
[EMAIL PROTECTED] tel:+64(9)373-7599x86730 fax:+64(9)373-7090
www.health.auckland.ac.nz/pharmacology/staff/nholford
_______________________________________________________________________________________________________________________________