A0216
Title: Optimal design for mixed effects models
Authors: John Stufken - George Mason University (United States) [presenting]
Abstract: Identifying optimal designs for correlated data is a difficult problem. Many classical results for independent data have no obvious generalization to correlated data. We propose a method to identify locally optimal designs for classes of linear, generalized linear, and nonlinear mixed effects models under commonly used optimality criteria by extending results for independent data. We demonstrate the method through a real life study, and investigate robustness of design efficiency to mis-specification of the covariance matrix for the random effects.
