American Journal of Epidemiology Practice of Epidemiology Statistical Criteria for Selecting the Optimal Number of Untreated Subjects Matched to Each Treated Subject When Using Many-to-one Matching on the Propensity Score