|
OPTIMUM DESIGNS FOR DISCRIMINATION BETWEEN TWO NONLINEAR MULTIVARIATE DYNAMIC MIXED-EFFECTS MODELS Bartosz Kuczewski1,2, Barbara Bogacka1, Dariusz Uciński2 1Queen Mary, University of London, School of Mathematical Sciences, Mile End Road, London E1 4NS, UK, e-mail: b.bogacka@qmul.ac.uk 2University of Zielona Góra, Institute of Control and Computation Engineering, ul. Podgórna 50, 65-246 Zielona Góra, Poland |
The paper concerns a problem of finding powerful experimental designs in order to discriminate between two alternative nonlinear multivariate dynamic mixed-effects statistical models. The T-optimality criterion developed for fixed models with heteroscedastic errors is generalized and used after linearization of the candidate models by Taylor series expansion around the mean value of the parameters. The relevant equivalence theorem is proved. A numerical algorithm for finding optimal designs based on a Wynn-type iterative procedure is constructed. T-optimum designs for discrimination between two pharmacokinetic multiresponse models are calculated as an example.
model selection, T-optimum design, pharmacokinetic models, random parameters
