Toward Responsible Learning Analytics in Special Education: An Exploratory Logistic Regression Analysis of IEP Goal Completion in India

Background: Learning analytics may support individualized planning for students with disabilities, but predictive claims require data provenance, outcome definitions, validation, and careful handling of rare outcomes. Objective: To examine factors associated with recorded Individualized Education Program (IEP) goal completion in an Indian special-education dataset and assess limitations of the fitted logistic model as a predictive tool. Methods: This exploratory analysis used 100 identified student records from government, private, and NGO-run settings. Seven students (7%) had IEP goal completion. Descriptive statistics, an exact test for parent engagement, and logistic regression using age, attendance, and pre-intervention score were reported. No train-test split, cross-validation, external validation, or machine-learning comparison was available; results were interpreted as exploratory associations. Results: Pre-intervention score was associated with IEP goal completion (B=0.178, p=.020; OR=1.195, 95% CI 1.028-1.389). Age and attendance were not statistically significant. Although apparent accuracy was 93%, only one of seven cases was correctly classified, yielding sensitivity of 14.3% and balanced accuracy of 56.6%. Conclusions: Baseline performance may warrant further study as a candidate predictor, but class imbalance and absent validation preclude claims of deployable prediction or AI effectiveness. Prospectively documented, multi-site datasets are required before predictive analytics can guide IEP decisions.