Regression models estimated after matching control and treatment groups on observables must explicitly account for the matching step in order to estimate asymptotically valid standard errors.

Menu