Are propensity scores really superior to standard multivariable analysis?