Comparison of the Accuracy of Item Response Theory Models in Estimating Student’s Ability

This study aims to determine the item response theory model which is more accurate in estimating students' mathematical abilities. The models compared in this study are Multiple Choice Model and Three-Parameter Logistic Model. Data used in this study are the responses of a mathematical test of 1704 eighth-grade junior high school students from six schools in the Depok City, West Java. The Sampling is done by using a purposive random sampling technique. The mathematics test used for research data collection consisted of 30 multiple choice format items. After the data is obtained, Research hypotheses were tested using the variance test method (F-test) to find out which model is more accurate in estimating ability parameters. The results showed that Fvalue is obtained 1.089, and  Ftable is 1.087, the value of Fvalue > Ftable, so it concluded that Ho rejected. That means Multiple Choice Model is more accurate than Three-Parameter Logistic Model in estimating the parameters of students' mathematical abilities. This makes the Multiple-Choice Model a recommended model for estimating mathematical ability in MC item format tests, especially in the field of mathematics and other fields that have similar characteristics.