Bayesian structural equation modelling of associations between digital access and data literacy competencies in South African university students

This study examines the association between reliable internet access during high school and university students’ self-assessed skills in data analysis and visualisation, amidst the growing emphasis on data literacy in higher education. It also examines the mediating role of data analysis competence and the robustness of these associations across different statistical models. The research employed the Bayesian structural equation model to analyse survey data from 413 first-year students at a South African university. The model incorporated latent variables for data analysis and visualisation, while controlling for gender, age, academic discipline, and mathematical background. Measurement reliability, modelled pathway estimates, and robustness were evaluated using posterior predictive checks, sensitivity analyses, and leave-one-out cross-validation. Reliable internet use during high school was positively associated with university-level data analysis competence (beta = 0.234, 95% CrI [0.089, 0.378]), with a small-to-medium standardised association (d = 0.328). Data analysis competence formed a statistically supported indirect pathway linking digital access and visualisation skills (indirect pathway estimate = 0.146), accounting for roughly 63% of the total association. The results were consistent across various priors and error distributions. Covariate analysis showed significant subgroup disparities. Because the study relies on cross-sectional survey data and retrospective reports of high-school internet access, these findings are interpreted as statistically supported associations consistent with the proposed mediation model rather than as identified causal effects. Educational policy may benefit from pairing infrastructure expansion with skill-building support as the observed association pattern is consistent with the importance of both access and competency development.