AN OPTIMAL, UNBIASED CLASSIFICATION RULE FOR MASTERY TESTING BASED ONLONGITUDINAL DATA

An optimal, unbiased classification rule is proposed based on a longit udinal model for the measurement of change in ability. The proposed me thodology can be used as an additional tool for the year-to-year evalu ation of student progress as well as consideration of the master testi ng problem. In general, it predicts future level of knowledge by using information about level of knowledge at entrance, its rate of growth, and the amount of within-individual variation. An illustration shows how the individual-oriented threshold value above which a student can be considered a master depends on the intra-test score variation and h ence differs from student to student. Furthermore, it appears that inf ormation about growth of knowledge in the first year substantially imp roved the prediction of relative position of ability in the future.