Proposing a New IRT (Item Response Theory) Guessing Model That Adjusts Both Ability and Difficulty

• In Event: Roundtable Session 117
  In Roundtable Session: Item Response Theory

Abstract

This research proposed a new guessing model that ability and difficulty were both adjusted, describing the response as a result of both ability and guessing. A Bayesian MCMC simulation study will be conducted to evaluate the estimation of the new guessing model, comparing with 2-PL and 3-PL IRT models. Furthermore, a real data analysis will be followed to investigate the model fit. A preliminary real data analysis showed that the new ability and difficulty adjusted guessing model was the best compare with 2-PL, 3-PL, and 3-PL AG (Martin, del Pino, and De Boeck, 2006) models.

Authors

• Yao Wen, University of Wisconsin - Milwaukee

• Bo Zhang, University of Wisconsin - Milwaukee