A0515
Title: Sequential designs for possibly misspecified mixed ANCOVA models with longitudinal data
Authors: Sanjoy Sinha - Carleton University (Canada)
Xiaojian Xu - Brock University (Canada) [presenting]
Abstract: The construction of robust sequential designs are studied for linear mixed models applied to longitudinal or clustered data, where both treatment and covariate effects are present. The objective is to determine optimal allocations of sample units to treatments, as well as appropriate covariate levels, while accounting for possible misspecification in the assumed model structure. The design approach balances two goals: Minimizing the determinant of the estimated variance-covariance matrix to improve precision, and reducing potential estimation/prediction bias arising from model misspecification. The empirical performance of the proposed designs is also examined through simulation studies.
