B1406
Title: Robust designs in mixed ANCOVA models for clustered data
Authors: Xiaojian Xu - Brock University (Canada) [presenting]
Sanjoy Sinha - Carleton University (Canada)
Abstract: The construction of robust designs is discussed for the linear mixed model with clustered outcomes or repeated measurements. Both treatment and covariate effects are involved in the mixed model. A goal of the design scheme is to determine the proportions of sample units allocated to each treatment as well as the covariate levels for the treatments with awareness of possible misspecifications in the assumed models. In addition, a sequential design process is also proposed for a longitudinal study to select the follow-up time points optimally. The resulting designs are obtained by minimizing the determinant of the estimated variance-covariance matrix alone and reducing the possible estimation biases due to the model imprecision. The empirical properties of the proposed designs are also studied through simulations.