Penskoran yang Fair pada Tes Matematika Berbentuk Pilihan Ganda Menggunakan Item Response Theory

Abstract

The scoring of multiple choice test items with the conventional model causes disparities in the interpretation of the scoring results. This causes unfair scoring, because the scoring does not pay attention to the characteristics of the items attached to the test. To overcome this problem, the purpose of writing this article is to apply the IRT model to make scoring more fair, so that multiple choice tests are able to provide a more precise and objective interpretation of scores. The multiple choice test used in this study is a math test on Algebra content of 15 items. The test was piloted on 65 grade 8 junior high school students. This data was used to calibrate test items using IRT. Model fit is determined by the M2 method. The instrument is compatible with the Rasch, 2PL and 3PL models, but only the Rasch and 2PL models used in this article. The number of items that fit in both models is the same as the number of multiple choice test items. Furthermore, the item parameters are used as the basis for scoring. The comparison results show that the IRT model is able to provide a more objective and precise scoring, based on the characteristics possessed by the test items. Model 2 PL is able to provide the most fair scoring than the Rasch and conventional models. The characteristics of the discriminant parameters in the 2PL model make it possible to distinguish students' abilities. especially for students who have the same number of correct answers, but the pattern of answers is different