Lessons learned from DIF algebra problems

Abstract

In light of the low representation of women in high-level technological jobs, many studies have addressed the issue of gender differences in mathematics. This study investigated those differences from the angle of gender-related differential item functioning (DIF) on algebra and algebra-related items on the Florida Comprehensive Assessment Test (FCAT) and on the 2003 Trends in International Mathematics and Science Study (TIMSS). The goal of the study was to identify the characteristics of items that contribute to DIF, finding differences for female and male students whose total scores on the test matched. More than 300 items from 3 years of the FCAT, Grades 8, 9, and 10, were coded according to the mathematical content, context topic, and other mathematical and nonmathematical characteristics. The Mantel-Haenszel and standardization procedures were used to quantify DIF. Two content categories, geometrical measurement and informal algebra, favored male students. The algebraic manipulations category favored female students. Several context topics were found to contribute to DIF: Recreational topics favored male students, and social studies topics favored female students. Items that required providing an approximate answer were challenging for female students. Items involving converting units favored male students, and items with noncomputed answers favored female students. Characteristics contributing to DIF on the FCAT were compared with characteristics of DIF items from the TIMSS. Data from the TIMSS were analyzed for U.S. eighth graders. DIF items were identified and characteristics of released items were compared to those on the FCAT. Findings on content categories were confirmed. In addition, items that tested concepts of fractions were common among the DIF items favoring male students. Items with patterns were common among the DIF items favoring female students. The topic of the context did not benefit either gender, although female students had a high proportion of no-context DIF items. The results suggested that there are patterns of differences in mathematics performance for male and female students who presumably have the same ability. The results also indicated that differential functioning should be investigated on the mathematics concept level in addition to the item level to study performance of different demographic or latent groups.