Differential item functioning analysis of Chile, Costa Rica, and Mexico's PISA 2015 science items

The purpose of this research study is to identify items with culture-related and/or gender-related differential item functioning (DIF), and examine the level of DIF manifested among a set of binary-scored science items from PISA 2015. The data pertaining to 16 dichotomous PISA 2015 science items from 2828 students from three Latin American countries was used to run the DIF analysis through Mantel- Haenszel (MH) and Logistic Regression (LR) procedures across culture and gender groups. Subsequently, the uniform DIF results provided by both techniques were analyzed and DIF effect size was calculated based on the corresponding classification indexes for MH and LR. When the variable of culture was examined, the findings from this investigation revealed 11 items with large DIF (C level): 4 for Chile and Costa Rica, 3 for Chile and Mexico, and 4 for Costa Rica and Mexico. It was found that Costa Rica was more favored than Chile; Chile was more favored than Mexico; yet Costa Rica and Mexico were equally favored. This study detected 3 items with C level DIF due to gender differences. It was found that for Chile the only item detected favored male students, whereas for Costa Rica 1 item favored males and 1 item favored females. After completing this investigation it was concluded that the analyzed set of 16 binary-scored PISA 2015 science items functioned more equally across genders than across cultures.