Split Samples and Design Sensitivity in Observational Studies

Abstract
An observational or nonrandomized study of treatment effects may be biased by failure to control for some relevant covariate that was not measured. The design of an observational study is known to ...