Abstract:
This study aimed at investigating differential item functioning (DIF) on the Special English Test of the Iranian National University Entrance Exam (INUEE). The effect of gender and subject area was taken into account. The study utilized one-parameter IRT model with a sample of 36000 students who sat for the INUEE Special English Test in 2004 and/or 2005. The findings confirmed the presence of DIF on this test. The cloze test indicated the lowest DIF whereas language function indicated the highest DIF. The results also revealed some general gender DIF patterns across the subject area. Females were favored on the three sections of grammar، language function، and the cloze test، whereas males were favored on the vocabulary and word order sections. The reading comprehension section favored males and females equally. It was also concluded that the item format alone could not explain DIF. In other words، it is the subject area or the interaction of the subject area and item format that determines the degree and direction of DIF.
Machine summary:
com Abstract This study aimed at investigating differential item functioning (DIF) on the Special English Test of the Iranian National University Entrance Exam (INUEE).
Studies of this type generally suggest that reading comprehension is differentially easier for females, whereas males are better performers on antonyms and analogies and that women are likely to perform better on items with more contexts (Carlton & Harris, 1992); that females are much superior in writing and speaking though in reading and vocabulary the difference is not significant, and that males are superior in listening comprehension (Cole, 1997); that females are favored in grammar and vocabulary whereas males in cloze tests (Lin & Wu, 2003); and that items classified as Mood/Impression/Tone are easier for females whereas items classified as Logical Inference are more likely to favor males regardless of item content (Pae, 2004).
The present study was also basically an observational one focusing on the specification of the number and type of DIF items in different sections of a high-stakes test in Iran (INUEE).
This along with the fact that separating the sections of a test for DIF analysis could lead to more accurate results and therefore reduce the probability of type 1 error which mistakenly highlights items for DIF (Clauser, & Mazor, 1998; Donlon, Hicks, & Wallmark, 1980; Le, 2000 and Reeve 2003) led the researchers to break the test into different sections.
Gender DIF in the English test: Number of items favoring each group Grammar Vocabulary Word order Language function Cloze Reading Total The above table indicates that the cloze section carried the lowest amount of DIF whereas the language function carried the highest.