Multiple-objective designs in a dose-response experiment

This article is an extension of the work of Huang and Wong (1), who found dual-objective designs for models with a continuous outcome. We consider quantal dose-response experiments with a binary outcome and develop multiple-objective designs for two or more Bayesian optimality criteria. Using the logit model as an illustrative example, we construct numerically optimal designs for estimating model parameters and percentiles, with possibly unequal interest in each of the objectives. We also show that the popular equal dosage assignment rule can be a rather inefficient design for estimating model parameters under a Bayesian setup.