Bootstrapping is often used as a substitute for asymptotic distributions when the latter are not available. Recent developments in the theory of the bootstrap show that combining the bootstrap with a known asymptotic distribution yields inferences that improve on those drawn from asymptotic...