COMPUTERIZED ADAPTIVE TEST (CAT) OPTIMIZATION USING THE EFFICIENCY BALANCED INFORMATION (EBI) METHOD

Abstract

This study aims to optimize Efficiency Balanced Information (EBI) for selecting items in the computer-based adaptive exam model (CAT), usually used in education. This optimization is used to reduce problems that often occur in CAT based on item response theory (TRB), namely the emergence of ability bias, large Mean Absolute Error (MAE), too many test lengths, and uncontrolled Standard Error Estimation (SEE). The success of the CAT-based exam is strongly influenced by the item selection method; therefore, an appropriate item selection strategy is needed. This study proposes selecting items using the EBI method to solve this problem. The simulation results show that the grain bias is slightly around zero, and CMAE is consistent in the allowable theta area. The average number of items presented is less than 20 when -2<θ<2. The average test length increases to 25 seconds | θ | > 2. The items presented show the different power parameters of questions (a) at the beginning of the test more than at the end of the test; therefore, this method can be the right solution for optimizing accuracy in selecting items