Equating Five Sets of Junior High School National Exam Questions for the Mathematics Subject

This research aims to determine the characteristics of test items and equate the National Examination test items for junior high school students in Mathematics for the 2015 academic year. The study used Item Response Theory (IRT), which ensures that students’ ability level is not dependent on the items’ difficulty level, and the items’ difficulty level is independent of the students taking the test. This study follows an exploratory descriptive approach, describing the characteristics and equivalence of the test packages created in five different versions. Data were sourced from the results of the 2015 National Examination in Mathematics for junior high school students, consisting of 40 items from 5000 students. The quantitative analysis employed the 2PL model to examine item difficulty, discrimination power, and model fit using the IRT approach through Bilog MG version 3.0. The results identified 10 questions with the same indicators that met the criteria for good items, which were used as anchors. For equating, the mean-ratio, mean-sigma, Haebara, and Stocking Lord methods were used with the R program. The equated packages were package 3 and 5, using the Haebara method with the equation b∗3 = 1, 036783b3 + 0, 027869 yielding an RMSD (b) = 0.00995 and RMSD (a) = 0.08968, while package 4 and package 5 were equated using the mean-ratio method with the equation b∗3 = 0, 964830b3 − 0, 018938 with an RMSD (b) = 0.01769 and an RMSD (a) = 0.08211.