Applying the item response theory with two-parameter, three-parameter models in the evaluation of multiple choice tests

Abstract

The article presents the results of analyzing and evaluating multiple-choice items based on Item Response Theory (IRT) with two-parameter and three-parameter models through analysis results of data from R software (package ltm). Data in this study are the results of answering 50 multiple-choice items of 590 students who took the English 1 test organized at Dong Thap University in 2018. By evaluating each multiple-choice item based on their difficulty, discrimination parameters and guessing parameter according to the models, the study has identified good items to put into item bank, and point out items that are not really optimal, thus should continue to be considered before being put into use. The review and analysis of multiple-choice items based on both models help evaluate items more comprehensive and item selection more accurate. In addition, the research results show that if the evaluation of the test is only based on the subjective opinions of professional lecturers, not on the process of analyzing and evaluating based on IRT, the not good items could be introduced into the test without being detected.