For objective causal inference, design trumps analysis

Donald B. Rubin

For obtaining causal inferences that are objective, and therefore have the best chance of revealing scientific truths, carefully designed and executed randomized experiments are ge…