A0270
Title: Balanced and robust randomized treatment assignments: The finite selection model
Authors: Ambarish Chattopadhyay - Stanford University (United States)
Carl Morris - Harvard University (United States)
Jose Zubizarreta - Harvard University (United States) [presenting]
Abstract: The Finite Selection Model (FSM) was developed previously for the design of the RAND Health Insurance Experiment (HIE), one of the largest and most comprehensive social science experiments conducted in the U.S. In the FSM, a treatment group at each of its turns selects the available unit that maximally improves the combined quality of its resulting group of units according to a common optimality criterion. In the HIE and beyond, the FSM is revisited, formalized, and extended as a general tool for experimental design. Leveraging the idea of D-optimality, a new selection criterion in the FSM is proposed and analyzed. The FSM using the D-optimal selection function has no tuning parameters, is affine invariant, and can retrieve classical designs such as randomized block and matched-pair designs. For multi-arm experiments, algorithms are proposed to generate a selection order of treatments. FSM's performance is demonstrated in a case study based on the HIE, a simulation study, and in ten randomized studies from the health and social sciences. On average, the FSM achieves 68\% and 56\% better covariate balance than complete randomization and rerandomization in a typical study. The FSM is recommended to be considered in experimental design for its conceptual simplicity, efficiency, and robustness.
