Levi on the Dogma of Randomization in Experiments

Abstract

It is my goal in this paper to apply Isaac Levi's theory, as it is developed in his latest book The Enterprise of Knowledge, to appraise the well entrenched methodological doctrine of randomized experimental designs.1 I hope to argue that whether we understand the conclusion of statistical 'inference', based on data from a randomized design, to be a statement of evidential support, e.g. a credal probability, or to be a statistical decision, in either case the familiar argument in defense of randomization as a protection against inaccurate 'inference' due to a biased sample is invalid.