A Comparison of Several Methods for Assessing Student Mastery in Objectives-Based Instuctional Programs

Abstract

In objectives-based instructional programs where relatively short criterion-referenced tests are administered to estimate student mastery for the purpose of monitoring a student through the program, estimates which maximally utilize the information that can be obtained from the student during the alloted testing time are required. Bayesian estimates, which utilize prior information about the student, direct information provided by the student, and collateral information in the test data of other students, appear to be ideally suited for this purpose. In this paper, the relative merits of several methods, Bayesian and classical, for the estimation of student student mastery are investigated. The effects of such factors as group homogeneity, test length, sample size, and prior information on the accuracy of the estimates as well as on decision-making accuracy are studied through computer simulations. It is shown that certain Bayesian estimates are superior to others, and the implications of the findings for objectives-based instructional programs are discussed.