The Ability of Item Parameters as Predictors of Item Fit Index According to Item Response Theory (IRT)

Abstract

This study aimed at inspecting the ability of item parameters difficulty, discrimination, and guessing as predictors of item fit index () according to item response theory (IRT). To achieve the goals of study the researcher used an achievement test in statistics consist of (29) four multiple choice items applied to sample consists of)507 (female students from Faculty of Education. SPSS package used to verify assumptions of IRT models, then BILOG-MG used to extract item parameters and item fit index for three parameters model, and identify fitting items, then extract correlation coefficients for item parameters and item fit index, finally estimate the regression line for them. Results showed that the number of fitted items varied according to number of ability intervals and was the best in case of six intervals, also item discrimination had a significant relationship with item fit index, the regression model predicted item fit index using transformed discrimination and difficulty parameters with superiority for six intervals case. Finally results discussed in light of literature review and it was recommended to study item fit for different sample sizes, fit indices, statistical models and item types