STOCHASTIC OPTIMIZATION ALGORITHMS OF A BAYESIAN DESIGN CRITERION FORBAYESIAN PARAMETER-ESTIMATION OF NONLINEAR-REGRESSION MODELS - APPLICATION IN PHARMACOKINETICS

This article proposes three stochastic algorithms to optimize a Bayesi an design criterion for Bayesian estimation of the parameters of nonli near regression models; this criterion is the information expected fro m an experiment. The first algorithm is based on a stochastic version of the simplex with an adaptive sampling procedure. The others are sto chastic approximation algorithms: the Kiefer-Wolfowitz and the pseudog radient algorithms. We first present the information criterion and the optimization algorithms. The efficiency of each algorithm for optimiz ing this Bayesian design criterion is then assessed by a simulation st udy for a nonlinear model assuming a discrete prior distribution. An a pplication for designing an experiment to estimate the kinetics of rad ioiodine thyroid uptake is then proposed. (C) 1997 Elsevier Science In c.