Understanding Students' Subjective and Eudaimonic Well-Being: Combining both Machine Learning and Classical Statistics

There is a vast literature focusing on students' learning and academic achievement. However, less research has been conducted to explore factors that contribute to student well-being. Rooted in the ecological framework, this study aimed to compare the relative importance of the individual-, microsystem-, and mesosystem-level factors in predicting students' subjective and eudaimonic well-being. Hong Kong data from the Programme for International Student Assessment (PISA) 2018 involving 6,037 students were analyzed. Machine learning (i.e., random forest algorithm) was used to identify the most powerful predictors of well-being. This was then followed by hierarchical linear modelling to examine the parameter estimates and account for the nested structure of the data. Results showed that four variables were the most important predictors of subjective well-being: students' sense of belonging to the school, parents' emotional support, resilience, and general fear of failure. For eudaimonic well-being, resilience, mastery goal orientation, and work mastery were the most important predictors. Theoretical and practical implications are discussed.