Assessing Dimensionality in Compensatory Multidimensional Item Response Theory Models With Complex Structure

• In Event: Item Response Theory: Applications and Extensions

Abstract

This simulation study investigates the effect of complex structure on dimensionality assessment in compensatory multidimensional item response theory (MIRT) modeled data using dimensionality assessment procedures based on conditional covariances (DETECT) and factor analytical approach (NOHARM). The DETECT-based methods typically outperformed the NOHARM-based methods in correct identification of dimensionality; especially when correlations were ≤ .60, data exhibited ≤ 30% complexity, and larger N. As the complexity increased and the sample size decreased, the performance of the methods typically diminished. As the complexity increased, it also became more difficult to label the resulting sets of items from DETECT as dimension-like. DETECT was more consistent in classification of factorially simple than complex items. NOHARM-based methods of χ2G/D and ALR generally outperformed RMSR.

Authors

• Dubravka Svetina, Indiana University - Bloomington

• Roy Levy, Arizona State University