Psychometric Analysis of Mathematics Achievement Test Using Item Response Theory

Abstract

This study determined the psychometric properties of the Mathematics Achievement Test (MAT) using Item Response Theory (IRT). Three popular IRT models namely, 1PL, 2Pl, and 3PL IRT models were utilized following unidimensionality test. Data from 2448 second year high school students from selected public and private secondary schools in Region IVA were the subjects for analysis . Analyses of data were performed in R. Results indicated that the entire 50-item cognitive test did not meet the unidimensionality assumption. Items were grouped according to content strands and were subjected again to dimensionality test using Modified Parallel Analysis. The remaining items after some deletion process were subjected to item calibration. Data fit analysis were performed to each strand . The 1PL, 2PL, and 3PL IRT models fit the different strands of the MAT reasonably well. The researchers established an item pool that can be used in estimating students’ mathematics ability