Abstract
The analysis of differential item functioning (DIF) examines whether item responses differ according to characteristics such as language and ethnicity, when people with matching ability levels respond differently to the items. This analysis can be performed by calculating various statistics, one of the most important being the Mantel-Haenszel, which can be carried out with software programs such as EZDIF, DIFAS and, more recently, EASY-DIF. In this context, the aim of the present study is to compare these three software programs by using simulated and real data. The procedural characteristics and the results obtained from the same dataset were thus compared by the three programs. DIFAS and EASY-DIF always provide equivalent results, while EZDIF is less accurate when using the thin matching strategy. The results also showed that DIFAS and EASY-DIF were the easiest to run, especially for testing practitioners, with the second offering a broader range of results for key characteristics for detecting DIF.