Multidimensional Dataset on Mathematics Performance in Tanzanian Secondary Schools (2025)

The dataset used in this study was constructed from primary data collected through a structured questionnaire administered to Form Three students in 2025. The target respondents were students who had completed the Form Two National Assessment (FTNA) in 2024 and had official examination results. The questionnaire was designed to capture retrospective information on factors influencing students’ learning environments during their Form Two academic year, enabling linkage between these factors and their subsequent mathematics performance in FTNA. Data were collected from five regions representing major geographical zones of Tanzania: Dar es Salaam Region (Coastal Zone), Arusha Region (Northern Zone), Dodoma Region (Central Zone), Mwanza Region (Lake Zone), and Mbeya Region (Southern Highlands Zone). To ensure validity, the instrument underwent content validation through expert review and alignment with established theoretical and empirical frameworks. It comprised multiple sections covering demographic characteristics, household assets, instructional practices, student attitudes and perceptions toward mathematics, family structure, and key psychometric constructs such as mathematics confidence and career utility. In total, 106 variables were collected, providing a high-dimensional dataset suitable for advanced machine learning analysis. Prior to full deployment, the questionnaire was piloted among a small group of students at Kisasa Secondary School in Dodoma Region to refine clarity and reliability. The dependent variable (students’ mathematics performance) was operationalized using FTNA results, measured on an ordinal scale ranging from A (Excellent) to F (Fail). To ensure data accuracy and authenticity, students’ reported examination numbers were used to retrieve and verify their official results from the National Examinations Council of Tanzania (NECTA) online results portal. This integration of self-reported and verified administrative data enhances the reliability and robustness of the dataset for subsequent analysis. Version update note: The grade coding of the dependent variable (FTNA_Math_Grade) has been revised in this version to adopt an intuitive ascending scale in which higher values represent better performance: F (Fail) = 1, D (Satisfactory) = 2, C (Good) = 3, B (Very Good) = 4, and A (Excellent) = 5. This update replaces the descending encoding used in the previous version (A=1 to F=5) and ensures consistency with other manuscript associated with this dataset.