The 2015 Trends in International Mathematics and Science Study (TIMSS) involved 57 countries and 43 different languages to assess students’ achievement in mathematics and science. The purpose of this study is to evaluate whether items and test scores are affected as the differences between language families and cultures increase. Using differential item functioning (DIF) procedures, we compared the consistency of students’ performance across three combinations of languages and countries: (a) same language but different countries, (b) same countries but different languages, and (c) different languages and different countries. The analyses consisted of the detection of the number of DIF items for all paired comparisons within each condition, the direction of DIF, the magnitude of DIF, and the differences between test characteristic curves. As the countries were more distant with respect to cultures and language families, the presence of DIF increased. The magnitude of DIF was greatest when both language and country differed, and smallest when the languages were same, but the countries were different. Results suggest that when TIMSS results are compared across countries, the language- and country-specific differences which could reflect cultural, curriculum, or other differences should be considered.
When didactics meet data science: process data analysis in large-scale mathematics assessment in France
