A model from pharmacokinetics;
A comparison with NLME



ADMB Files
Code: pheno.tpl
Data: pheno.dat
Initial values: pheno.pinExpected Results: pheno.par

Model description

Pinheiro & Bates (2000, Ch. 6.4) fitted a so-called 'one-compartment open model' to a dataset known as the 'phenopharbital data'. A patient is given a dose D at time td, giving rise to a phenopharbital concentration ct at a later time t:
ct = D/V exp[-Cl/V(t-td)],

where V and Cl are parameters (the so-called 'Volume of concentration' and the 'Clearance', respectively). Doses given at different time points contribute additively to ct.
Pinheiro & Bates (2000) fitted a model with a linear predictor (and a log-link) for each of the paramere V and Cl. Each of the linear predictors contained one covariate Wt and one random effect. A full description of the model can be found here: pheno.pdf

Results

The results produced by ADMB-RE and nlme are rather similar. The small difference is caused by the fact that the two methods use different approximations to the likelihood function. (ADMB-RE uses the Laplace approximation, and for nlme the reader is referred to ( Pinheiro & Bates, 2000, Ch. 7).)