Mixed effects models, which allow the model parameters to vary randomly
between blocks, are commonly used in pharmacometrics when modeling
pharmacokinetic and pharmacodynamic data. In these cases, each patient
is treated as a ‘block’ so that parameters such as the rate at which
the body clears a drug may vary from patient to patient.
This talk outlines some approaches to designing experiments for such
models when the aim is to minimise the confidence region of parameter
estimates. Such approaches are based on the information matrix, the
calculation of which becomes rather perilous for nonlinear and generalised
linear models when random effects are introduced.
This is joint work with John Eccleston and Stephen Duffull at the
University of Queensland, Australia.
Meet the speaker in Room 212 Cockins Hall at 4:30
p.m. Refreshments will be served.
