The Best IRT Logistic Model for a Pre- and Post-Testing System with Previous Tests
Tomoko Fujita, Tokai University, Japan ; Shin-ichi Mayekawa, Tokyo Institute of Technology, Japan
Society for Information Technology & Teacher Education International Conference, in Nashville, Tennessee, USA ISBN 978-1-880094-84-6 Publisher: Association for the Advancement of Computing in Education (AACE), Chesapeake, VA
The purpose of this research is to demonstrate an easy and accurate pre-and post-English language testing system in a university language program with 5,000 students each year. Previously conducted test items were used for easy construction of the testing system, and three item response models, 2PL, 3PL, and 3PLcFix were tested to find the most appropriate model of IRT equating. Results show that the 3PLcFix model was the best fitting model and the 3PL was the worst. Secondly, the average gain in ability among the 694 students who took the listening pre-and post-tests were calculated with both the most and least fitting model. The differences in the results were clear; the average gain in student ability as calibrated with the 3PL model is 0.282 while that from the 3PLcFix is 0,432.
Fujita, T. & Mayekawa, S.i. (2011). The Best IRT Logistic Model for a Pre- and Post-Testing System with Previous Tests. In M. Koehler & P. Mishra (Eds.), Proceedings of SITE 2011--Society for Information Technology & Teacher Education International Conference (pp. 24-30). Nashville, Tennessee, USA: Association for the Advancement of Computing in Education (AACE).