The Role of Bayesian Concepts in Test Development and Test Interpretation

Abstract

Column Editor's Note. Among Robert L. Thorndike's many contributions is his use of Bayesian thinking in test construction and his dissemination of information about Bayesian concepts. Thorndike's ability to render complex technical material comprehensible to nonspecialists prompted your editor to invite him to develop a series of three articles.In the first, readers are introduced to central tenets of Bayesian reasoning and are led to appreciate that it is not an esoteric approach reserved for specialists; rather, it is closely related to common sense. In the second article, to appear in next month's issue, Thorndike sketches some Bayesian approaches to test development. In the final installment, which will appear in November, Thondike illustrates how Bayesian thinking can and should be used to interpret test scores. When we interpret a score, we usually have prior test or nontest information about the examinee. “We should neither ignore nor overvalue this prior information but should skillfully use it to govern our interpretation of the new data.It is hoped that this trilogy will not only provide some interpretive skill but will also enhance commonsense thinking in a wide variety of professional and personal applications.—Gerald S. Hanna