Balancing Information with Imbalanced Item Banks in Multidimensional CAT
Michael J. Culbertson · 2014
In multidimensional Computerized Adaptive Testing (CAT), if item banks provide more information on one dimension than others, most item selection algorithms will over-emphasize the information-rich dimension. This study evaluates two proposed selection criteria based on mutual information that are intended to equalize ability estimate precision across dimensions.