If correlation doesn't imply causation, what does?

Open science advocate Michael Nielson writes about how scientists can infer causation in situations where it's not possible to do a randomized controlled trial.