RE: a question about control file

From: "Jakob Ribbing" Jakob.Ribbing@farmbio.uu.se Subject: RE: [NMusers] a question about control file Date: Wed, May 25, 2005 11:25 am Dear Chang, A good guess at initial estimates is easier to make if the parameters that you estimate actually mean something. The current interpretation of TVCL or THETA(1) is the typical value of clearance for an individual which has been given a dose of 1 mg, weights 1 (kg?), is 1-year old of SEX=0 (and on no concomitant medication). It is very hard to tell what clearance this rather untypical individual might have. Center your covariates on the median to get rid of this problem. Be sure you understand which units your parameters have in order to get sensible initial estimates... Further, you might consider using a more mechanistic model for the covariate effects and enzyme induction. Using AMT to drive the induction (?) is probably not what you want to do since this variable is set to zero at the time-points where you have an observation in your data file. The induction on the other hand has a time delay. Good Luck! /Jakob _______________________________________________________