Item-Attribute Misspecifications and the Reparameterized Unified Model for Cognitive Diagnosis

• In Event: Poster Session 10
  In Poster Session: Cognition and Assessment SIG Poster Session

Abstract

The reduced reparameterized unified model (RUM) is a diagnostic classification model that uses dichotomous item responses to estimate a student’s mastery of latent attributes. The RUM allows every item in an instrument to behave differently and allows every attribute to affect the behavior of each item using that attribute. This paper proposes a simulation study of the RUM to investigate student classification performance when attributes are misclassified to items. This misclassification will be studied under various numbers of items, sample sizes, and item discriminations. Outcomes will include correct attribute classification rates as well as overall correct student classification rates.

Authors

• Robert John Sandbach, University of Florida

• M. David Miller, University of Florida