Consequences of the Proportional Constraint in Multilevel Multidimensional Item Response Theory Models

• In Event: 58.028 - Applications of Item Response Theory

Abstract

Three-level item response theory (IRT) models are used to account for cluster structures reflected in test data. Traditionally, the research on the consequences of ignoring cluster structures assumed that the structures were of proportional form (i.e., the item discriminations should be equal at Levels 2 [persons] and 3 [clusters] in the IRT model). Recently, research has shown that very different consequences arise when the ignored cluster structure is of nonproportional form (i.e., the item discriminations should be freely estimated at Levels 2 and 3). This study will fill the gap in the literature on the consequences of ignoring cluster structures of nonproportional form when the dimensional space at a given level is multidimensional (rather than just unidimensional).

Author

• Ken A. Fujimoto, Loyola University Chicago