A Hierarchical and Finite Mixture Modeling Approach to Mathematics Achievement: Evidence from PISA 2022

This study investigates the combined effects of cognitive, psychosocial, and cross-subject academic indicators on students’ mathematics ‎achievement across 38 countries that participated in the PISA 2022 creative thinking assessment. Drawing on data from 144,446 students, ‎we employed hierarchical linear modeling to examine how reading and science proficiency, engagement in creative activities (in and out of ‎school), perseverance, curiosity, and socioeconomic status (ESCS) predict mathematical performance. The results show that reading and ‎science are robust predictors of mathematics scores. ESCS and perseverance also demonstrated consistent positive effects, while creativity ‎showed context-specific associations, positive in some clusters and negative in others. To identify latent cross-national typologies, we ap-‎plied both K-means clustering and Gaussian Mixture Modeling (GMM) to country-level aggregates. Model comparison using the Bayesian ‎Information Criterion (BIC) favored the GMM solution, which was subsequently used to group countries for multigroup structural equation ‎modeling (MG-SEM). Results revealed significant variations in predictor effects across clusters, highlighting heterogeneity in pathways to ‎mathematics success. This study contributes to comparative education research by integrating hierarchical regression and latent classification ‎techniques, offering implications for instructional design and international education policy aimed at promoting mathematical literacy across ‎diverse systems‎.