Improving Subscale Estimation Using a Higher-Order Item Response Theory Model Combined With Adaptive Testing Algorithms

• In Event: Advances in Item Response Theory Estimation

Abstract

Two consortiums were formed in RTTT and they envisioned their next generation system to allow students to know whether they are on track for college and career readiness and allow parents to have clear and timely information about the progress of their children. Given these goals, it is of paramount importance to develop a psychometric model and delivery method that can produce valid and reliable scores for accountability purposes, along with a comprehensive diagnostic profile. This research proposes to use the higher order IRT model combined with hierarchical CAT algorithms. Four item selection algorithms are proposed and preliminary results show that all methods can generate more reliable overall and domain ability estimations.

Authors

• Chun Wang, University of Minnesota

• Hua-Hua Chang, University of Illinois at Urbana-Champaign

• Keith A. Boughton, CTB/McGraw-Hill LLC