RE: RE: Link different thetas to same omega using mu referencing
Thanks. You’re right. With my solution the eta is not constant for an
individual, instead it will vary for each value of the categorical covariate,
meaning that it is similar to IOV. With your solution we restrict only one eta
per individual not matter of the covariate value.
To exemplify this discussion one can think of two different oral formulations
with different absorption rate that are given in sequence. The question is then
if we would expect that a person who has “higher than normal” absorption rate
for the first formulation also would have a “higher than normal” absorption
rate for the second formulation. If the answer to that question is yes then one
should impose the restriction that you described.
//Jacob
Quoted reply history
From: Emmanuel Chigutsa [mailto:[email protected]]
Sent: den 13 april 2016 22:47
To: Leander, Jacob <[email protected]>;
[email protected]; 'Sadler, Brian' <[email protected]>;
[email protected]
Subject: Re: [NMusers] RE: Link different thetas to same omega using mu
referencing
Hi Jacob,
Yes, THETA(4) would represent the typical value (before exponentiation). In
summary, writing your covariate code after the definition of MU should
circumvent the time-varying problem. If you want everything on a log scale you
could try:
OCL=MU_1+ETA(1)
CL=EXP(OCL+CLCOV)
The issue with trying to use different etas for the different thetas as you had
was that only the variance (omega) was constrained to be same, but not
necessarily the same eta for each ID.
Mannie
________________________________
From: "Leander, Jacob"
<[email protected]<mailto:[email protected]>>
To: Emmanuel Chigutsa <[email protected]<mailto:[email protected]>>;
"[email protected]<mailto:[email protected]>"
<[email protected]<mailto:[email protected]>>; "'Sadler,
Brian'" <[email protected]<mailto:[email protected]>>;
"[email protected]<mailto:[email protected]>"
<[email protected]<mailto:[email protected]>>
Sent: Wednesday, April 13, 2016 2:00 PM
Subject: RE: [NMusers] RE: Link different thetas to same omega using mu
referencing
Thanks for the answers.
Just as Rupert pointed out it is not possible to model time-dependent
covariates in the MU expression. I am not sure what is happening if you do but
since the NONMEM manual explicitly tell you to avoid it you should not do it.
It seems that Mannies code instead uses a THETA(4) to model the “mean” value
and then use CLCOV to change the value for different covariates. Or what does
THETA(4) mean in your code Mannie? Remember that the problems was to impose
same omega element for different thetas depending on time-varying covariate and
also estimate all the parameters on a log scale.
//Jacob
From: [email protected]<mailto:[email protected]>
[mailto:[email protected]] On Behalf Of Emmanuel Chigutsa
Sent: den 13 april 2016 19:01
To: [email protected]<mailto:[email protected]>; 'Sadler,
Brian' <[email protected]<mailto:[email protected]>>;
[email protected]<mailto:[email protected]>
Subject: Re: [NMusers] RE: Link different thetas to same omega using mu
referencing
I think the following code should work:
IF(CAT.EQ.1) CLCOV = THETA(1)
IF(CAT.EQ.2) CLCOV = THETA(2)
IF(CAT.EQ.3) CLCOV = THETA(3)
MU_1 = THETA(4)
CL = EXP(MU_1+ETA(1))*(1+CLCOV)
$OMEGA BLOCK(1) 0.1
Mannie
________________________________
Confidentiality Notice: This message is private and may contain confidential
and proprietary information. If you have received this message in error, please
notify us and remove it from your system and note that you must not copy,
distribute or take any action in reliance on it. Any unauthorized use or
disclosure of the contents of this message is not permitted and may be unlawful.