Can students? attitudes towards mathematics and science be compared across countries? Evidence from measurement invariance modeling in TIMSS 2019

This study examines the measurement invariance and cross-country comparability of noncognitive scales in international educational large-scale assessments. Specifically, the TIMSS 2019 scales measuring fourth and eighth grade students' attitudes towards mathematics and science are analyzed using an item response theory (IRT) modeling approach with partial invariance assumptions. This approach allows the estimation of international (common) item parameters that are appropriate for most countries, while allowing for unique item parameter estimates in some countries to treat differential item functioning (DIF) and improve model-data fit. Results of this empirical evaluation show that items are comparable across most countries, supporting crosscultural comparisons of the mathematics and science attitudes scales.