Nonasymptotic Confidence Sequences for Sequential Estimation of Treatment Effects in Randomized Trials

Dr. Howard and his colleagues develop confidence sequences for sample average treatment effect estimation in randomized experiments. The sequences allow investigators to continuously monitor experiments and adaptively stop based on observed results. Building on time-uniform exponential concentration results, Dr. Howard arrives at a nonasymptotic coverage guarantee justified by the randomization mechanism in a nonparametric potential outcomes model. He will discuss this causal inference application and the underlying proof techniques, which are more generally applicable.

Speaker: Steve Howard, PhD, Researcher, The Voleon Group

Register: http://tiny.cc/EpiBioEmailList