Estimating Causal Effects: Inverse Probability Weighting