A stepwise test characteristic curve method to detect item parameter drift

item parameter drift;stepwise selection;test characteristic curve;item response theory;true score equating

Guo, Rui ; Chang ; Hua-Hua

关键词: item parameter drift;    stepwise selection;    test characteristic curve;    item response theory;    true score equating;   

Others  :  https://www.ideals.illinois.edu/bitstream/handle/2142/49806/Rui_Guo.pdf?sequence=1&isAllowed=y

美国|英语

来源: The Illinois Digital Environment for Access to Learning and Scholarship

【 摘 要 】

An important assumption of item response theory (IRT) based equatingis that the item parameters should be invariant over different testingoccasions. Sometimes, however, item parameters do not remain invariantdue to factors other than sampling error, and this is termed itemparameter drift (IPD). Several methods have been proposed to detect drifteditems. However, most of the existing methods aim at detecting thedrift in individual items, which may not be ideal when only theoverall test characteristic curve (TCC) is of interest to theusers. One such occasion in common practice is IRT-based true scoreequating, where the goal is to create a conversion table to make thetwo TCCs as close as possible. This paper introduces a stepwise test characteristic curve (Stepwise TCC) method to dynamically detect item parameter drift based on TCC without requirement to set any critical values.Comparisons were made between thenew method and two commonly used existing methods under thethree-parameter logistic model. Results show that the new methodperformed well in IPD detection.