DETECTING CROSSING DIFFERENTIAL ITEM FUNCTIONING (CDIF): BASED ON ITEM RESPONSE THEORY

Abstract

The objective of this study was to find out the empirical data of the index of crossing Differential Item Functioning (DIF) and the difference of sensitivity among methods in detecting crossing DIF according to the Item Response Theory. The study used quasi experimental methods with 1 x 3 designs. The independent variables were Raju’s area measure, Lord’s chi-square, likelihood ratio test, and sample size, while the dependent variable was index of crossing DIF. Methods for detecting uniform DIF had been well developed, where as those for identifying the presence of crossing DIF had not. The three methods compared were Raju’s area measure, Lord’s chi-square, and likelihood ratio test. Factors manipulated were sample size, ability differences between two groups, percentage of DIF, and the examinee’s response. The results indicated that all methods were good for the detection of crossing DIF, but Raju’s area measure was the most sensitive among those methods compared. Besides, there was no DIF caused by gender differences among the examinees with great ability.