Leveraging Causal Reasoning in Educational Data Mining: An Analysis of Brazilian Secondary Education.
This study combines EDM techniques with traditional models to address the lack of causal reasoning in previous EDM studies. Using large-scale Brazilian assessment data, unobserved confounders are mapped with causal graphs. A two-way logistic regression fixed effects model accounts for confounding factors, and predictive ability is evaluated. Classification rules and decision trees provide new insights, highlighting socio-economic factors, faculty education policies, and the crucial role of Brazilian states.
Manuscript publihsed on: Open Access