Matched Sampling for Causal Effects: Matching to Remove Bias in Observational Studies