The Problem Of Causality In Observational Data