Item Pool Designs for Cognitive Diagnosis Computerized Adaptive Testing (CD-CAT)

• In Event: 29.082 - Roundtable Session 9
  In Roundtable Session: 29.082-12 - Computer Adaptive Testing Designs and Analyses

Abstract

This paper investigates the impact of using different item pool designs on the performance of item selection algorithms in cognitive diagnosis computerized adaptive testing (CD-CAT). Three different item pool designs were considered in a simulation study. First, Design 1 consisted of items with all possible 2^K-1 q-vectors where K is the number of attributes. Second, Design 2 consisted of items with only a small number of attributes (e.g., one- or two-attribute items). Last, Design 3 consisted of items with only a large number of attributes (e.g., four- or five-attribute items). The results of this study showed that Design 1 and Design 2 resulted in similar mean test lengths; however, Design 3 yielded longer mean test lengths compared to the others.

Authors

• Mehmet Kaplan, Artvin Coruh University

• Cansu Ayan, Ankara University