A theoretical framework for mastery testingbased on item response theory and Bayesian decisiontheory is described. The idea of sequentialtesting is developed, with the goal of providingshorter tests for individuals who have clearlymastered (or clearly not mastered) a given subjectand longer tests for those individuals for whom the mastery decision is not as clear-cut. In a simulatedapplication of the approach to a professionalcertification examination, it is shown that averagetest lengths can be reduced by half without sacrificingclassification accuracy. Index terms: Bayesiandecision theory, computerized mastery testing, itemresponse theory, sequential testing, variable-length tests.
展开▼
