An Item Selection Method in Computerized Adaptive Testing for Cognitive Diagnosis

• In Event: Roundtable Session 16
  In Roundtable Session: Cognition and Assessment Roundtable

Abstract

This paper introduces a new item selection method, based on generalized deterministic inputs, noisy ``and" gate (G-DINA) model discrimination index that can be used in cognitive diagnosis computerized adaptive testing. The efficiency of the new method is compared with random, Kullback-Leibler information, and posterior-weighted-Kullback-Leibler item selection indices using a simulation study. In addition, the impact of item quality on attribute classification is also investigated. Reduced versions of the G-DINA model are used to generate data sets, and the item selection is performed under the G-DINA model. Two different test termination rules are considered. The results of the simulation study show that the new index outperformed the other indices, and it provided highest classification rates and shortest average test lengths.

Authors

• Mehmet Kaplan, Artvin Coruh University

• Jimmy de la Torre, Rutgers University