Hypothesized and Confounded Explanations in Theory Tests: A Bayesian Analysis

Abstract

Traditional views of research methodology hold that little, if any, useful information can be obtained from one or more confounded studies, unless the results from one study rule out or falsify an alternative explanation from a previous study. We present a Bayesian analysis of hypothesis testing to model knowledge accumulation from a series of confounded or unconfounded experiments. By applying this Bayesian analysis, we find that a hypothesis can receive support from a study with known flaws. Our analysis also implies that the status of an explanation is independent of whether it was proposed a priori or post hoc.