B1727
Title: Optimal experimental designs for efficiency of parameters and model discrimination
Authors: Saumen Mandal - University of Manitoba (Canada) [presenting]
Abstract: In optimal design, we generally assume that the statistical model is known at the design stage. However, in many situations, this is not the case. We need to implement a design that is efficient for two or more models, to discriminate between them, and select the best model. We consider different approaches to evaluate a criterion or a mixture of various criteria, and determine the efficiencies for the models and the parameters. We then consider a novel approach in which we optimize one objective subject to achieving a given efficiency for a parameter. We solve this constrained optimization problem by a Lagrangian approach. We eventually transform the problem to that of maximizing a number of functions simultaneously. These functions have a common maximum that is simultaneously attained at the optimum. We apply the method to some polynomial models, and discuss the results.
