Title: Applying IRT Growth Model to Investigate Item Parameter Drift on a Diagnostic Test for Toddler Development
Stream: Assessment and Learning Analytics
Presentation Type: Virtual Poster Presentation
Authors:
Chi-Chen Chen, National Academy for Educational Research, Taiwan
Bo-Hsien Hu, National Sun Yat-sen University, Taiwan
Abstract:
Item parameter drift (IPD) is a critical issue in educational measurement. Past researchers indicated that an item parameter may be misestimated if they function differently with time (Kopp & Jones, 2020; Park et al., 2016; Wells et al., 2002). As well known, biased parameters led to a lack of validity. In the literature, several methods had been proposed to identify the IPD. The Mantel-Haenszel (M-H) and area method were most used. The measurement of children's development is an essential issue in education. The students’ ability grows over time, and may interfere the accuracy of the IPD assessment. Unfortunately, both M-H and area methods did not consider the interference from the ability growth. Therefore, the authors develop a new method (G-IPD) to assess item parameter drift combining Embretson's (1991) IRT-based growth model. This study uses the Comprehensive Developmental Inventory for Infant and Toddlers from a longitudinal database. The interpersonal communication dimension (22 items) is used and assessed by the M-H, area, and G-IPD methods. The results indicated that 5, 10, and 8 items were shown IPD on the M-H, area, and G-IPD, respectively. The inconsistency results from the three methods reveals that different ways had their limitations. In a future study, the performance of G-IPD should be tested by a series of rigorous simulation studies. In addition, scales should be corrected by removing the IPD items or re-estimating the item parameters.
Conference Comments & Feedback
Place a comment using your LinkedIn profileShare this Presentation
Comments
Powered by WP LinkPress