Multidimensional Computerized Adaptive Testing for Ranking Items

• In Event: Roundtable Session 6
  In Roundtable Session: Computerized Testing: Applications and Advances

Abstract

Ranking items, where a set of statements are to be ranked, are widely used to measure interests, values, attitudes, or personality. In this study, we developed multidimensional computerized adaptive testing (MCAT) for ranking items and investigated two special issues that were seldom considered in standard MCAT studies: (a) computation time due to high dimensionally and large item pool, and (b) within-person statement exposure. To reduce computation time to a feasible length, we partitioned high-dimensional full item pool into a few low-dimensional sub-pools. To control within-person statement exposure, we implemented several control procedures. The effectiveness of these proposed methods was investigated by a series of simulations and found very effective.

Authors

• Chia-Wen Chen, The Hong Kong Institute of Education

• Wen-Chung Wang, The Hong Kong Institute of Education

• Sage Ro, IBM