The use of unidimensional parameter estimates of multidimensional items in adaptive testing.

Abstract

This study investigated the effect of using multidimensional items in a computerized adaptive test (CAT) setting which assumes that all items are unidimensional. Previous research has suggested that the composite of multidimensional abilities being estimated by a unidimensional IRT model is not constant throughout the entire unidimensional ability scale (Reckase, Carlson, Ackerman, & Spray, 1986). Results of this study suggest that unidimensional calibration of multidimensional data tends to "filter out" the multidimensionality. Items that measured a θ₁,θ₂ composite similar to the composite of the calibrated unidimensional θ scale had larger estimated unidimensional discrimination values. These items thus had a greater probability of being administered in a CAT where only the most informative items are selected. Results also suggest that if a CAT item pool contains items from several content areas measuring dissimilar θ₁,θ₂ composites, different unidimensional abilities may receive disparate proportions of items from the various content areas. Index terms: adaptive testing, item response theory, multidimensionality, parallel tests, test construction.