A-Stratified With B Blocking Methods Multistage Multidimensional Computerized Adaptive Testing to Control Item Exposure

• In Event: Roundtable Session 60
  In Roundtable Session: Computerized Adaptive Testing Issues

Abstract

The study explores the item exposure control issue in multidimensional computerized adaptive testing (MCAT). In MCAT, fisher information (FI) based D-optimality approach tends to pick up items with high discrimination parameters. The study investigates whether imposing a-stratified with b blocking (BAS) idea (Chang, Qian, & Ying, 2001) on FI (Weiss, 1982; Chang & Ying, 1999), Bayesian (Owen, 1975; Chang & Ying, 1999), and Kullback-Leibler information (KL; Chang & Ying, 1996) in MCAT could help control item exposure rate. In addition, a-stratified (AS; Chang & Ying, 1999) and Sympson-Hetter method (Sympson & Hetter, 1985) imposing on these methods are conducted for comparison. A simulation study assuming two-dimensional multidimensional three-parameter logistic model is conducted to explore whether improvement exists.

Authors

• Haiyan Lin, ACT, Inc.

• Hua-Hua Chang, University of Illinois at Urbana-Champaign