Re: [Fwd: CLIN PHAR STAT: Mixed Vs Fixed]

Date: Mon, 07 Aug 2000 23:35:18 +0200 From: Mats Karlsson <Mats.Karlsson@biof.uu.se> Subject: Re: [Fwd: CLIN PHAR STAT: Mixed Vs Fixed] Dear James and Nick, James alluded to a model for interaction between ETA's and EPS's. What he refers to is to put an interindividual random effect on the residual error. The reason you may want to do this is to acknowledge that residual error magnitude may vary between subjects (more or less model misspecification, better or worse compliers, etc). The way to do that in NONMEM is Y=F+EPS(1)*EXP(ETA(1)) where the magnitude of OM(1) would reflect the degree to which subjects differ in residual error magnitude. To use this option, you have to use FOCE INTERACTION. As opposed to interindividual variability in the ETA, which as James points out is tricky to estimate and relatively hard to grasp, interindividual variability in maginutde of residual error is often well estimated if you have a decent number of observations per subject (say >4). It can be a useful diagnostic (high POSTHOC residual variabiltiy in an individual-> poor fit) and it makes subjects whose profiles are not at all well described by the model to influence parameter estimates to a lower degree. On the topic of misspecification, I assume it would be possible also to estimate interindividual varaibility in auto-correlation between residuals, again providing a measure for varying degree of misspecification. I've never tried it or heard of anyone who has though. Best regards, Mats -- Mats Karlsson, PhD Professor of Biopharmaceutics and Pharmacokinetics Div. of Biopharmaceutics and Pharmacokinetics Dept of Pharmacy Faculty of Pharmacy Uppsala University Box 580 SE-751 23 Uppsala Sweden phone +46 18 471 4105 fax +46 18 471 4003 mats.karlsson@biof.uu.se