Accounting for Confounding in Observational Studies.