A Structural Equation Modelling Approach to Examination-Score Prediction

Understanding the determinants of students’ examination performance requires analytical approaches capable of capturing the complex and interrelated nature of academic, behavioral, and contextual factors. This study applies Structural Equation Modelling (SEM) to examine and predict students’ average examination scores using a multidimensional framework. Data comprising twelve observed variables were obtained from Kaggle and analyzed using SPSS AMOS. SEM was employed to simultaneously estimate direct and indirect relationships among academic behaviors, motivation, health status, parental support, socio-demographic characteristics, and exam performance. Model fit indices indicated an excellent fit to the data (RMSEA = 0.000, PCLOSE = 1.000). Results reveal that proximal academic and behavioral factors, particularly study habits, motivation, attendance percentage, study time, sleep hours, health status, and parental support are the strongest direct predictors of examination scores. In contrast, socio-demographic variables such as gender, parental education, and part-time employment showed no significant direct effects once behavioral and health-related factors were accounted for. Additionally, access to academic resources was found to influence exam performance indirectly through motivation. These findings underscore the importance of behavioral, psychological, and well-being factors over background characteristics in predicting academic success. The study demonstrates the utility of SEM as a robust and holistic tool for exam-score prediction and provides evidence-based insights to inform educational policy, institutional interventions, and student support strategies. Received: 14 January 2026 / Accepted: 28 February 2026 / Published: March 2026