AP-optimum Designs for Minimizing the Average Variance and Probability-Based Optimality

Abstract

The purpose of this paper is to introduce a new class of compound criteria and optimum designs that provide a specified balance between minimizing the average variance and high probability of a desired outcome. The proposed criterion called AP- optimality that combines A-optimality and P-optimality and address this issue for generalized linear models. An equivalence theorem for this criterion is provided and two numerical examples are presented for different GLMs to illustrate the achieved dual properties.