Groundhog Day: cause and effect and the primary importance of the finite population induced by randomization

The logical structure that enables randomized clinical trials to establish cause and effect is reviewed. Statisticians have a major role to define this structure clearly and to help clinicians apply statistical inference in ways directly related to their trials. Scientific importance of establishing limited cause and effect should be accented rather than inferences to generalized populations beyond the scope of a trial's actual random sampling. Authors of clinical reports should clearly define the randomization-induced populations to which their inferences apply. Careful attention to identifying this population of inference can lead to resolution of some issues commonly debated in the analysis of clinical trials. If formal inferences to generalized populations are attempted, these populations should also be carefully described, the assumed random sampling process should be carefully defined, and appropriate methods should be used that correspond with the assumed random sampling process.