Experimental design problem under mixed-effects non-linear models
by
Barbara Bogacka
School of Mathematical Sciences, Queen Mary, University of London, Mile End Road, E1 4NS London, UK
Coauthors: Maciej Patan

Many experiments are aimed at learning about average behaviour of some model parameters across a population of subjects. For example, in drug development studies, parameters such as drug absorbtion and drug elimination and their variability within the population are often of interest. Then designing an experiment for one individual in the population will not give the required information and a more general approach is needed. In our work we assume that the model parameters are random variables, and so they represent the population. This leads to the design problem under a mixed effects non-linear model.

In our talk we present a definition of an experimental design as a sampling plan for a population of individuals; we present some theoretical results for an optimum design for estimation of the population parameters and we discuss, via an example, various issues which occur in such design problems.

Date received: June 29, 2006