A technique that reweights observations to remove confounding bias by accounting for treatment assignment probabilities.