Rerandomization Methods to Minimize Interference in Experiments

• In Event: Challenges of Inference with Spillovers, Diffusion, and Interference

Abstract

We propose practical advice for minimizing interference at two points in the design of randomized experiments. In particular, we describe the benefits of rerandomization in unit selection and treatment assignment as strategies for creating buffers between units. We show how rerandomization in unit selection reduces the bias that standard causal estimators encounter when estimating causal effects in an interference-laden network.

Author

• Ryan T. Moore, American University