Improving Academic Performance in the Lowest Quartile: A Multilevel Analysis of a Remedial Intervention in Ecuadorian Public Schools

Vollständiger Abstract

Worum geht es in dieser Arbeit?

This study assessed the academic impact of an educational leveling intervention implemented across various institutions in Ecuador. Using a Difference-in-Differences (DiD) strategy within a multilevel modeling framework, we estimated changes in academic performance (GPA) over time between intervention and control groups. The analysis was conducted in RStudio using linear mixed-effects models (via the lmerTest package) to account for the hierarchical structure of the data (students nested within grades, institutions, and intervention types). Five models were compared, ranging from a full triple-interaction specification to main effects only, with Model 2 (including Time × Intervention and Time × Moment interactions) providing the best fit according to AICc weights and conditional R². The findings show a statistically significant relative improvement in GPA for the intervention group compared to the control group, particularly following the second implementation point. Moreover, performance differences were moderated by students' prior academic levels, highlighting the importance of stratified interventions. Diagnostic tests confirmed that assumptions of linearity, homoscedasticity, and independence were reasonably met. However, residuals were not perfectly normal; the literature supports the robustness of mixed models to such deviations, especially in large samples. These results suggest that targeted academic interventions can positively influence student achievement when they account for students' baseline performance levels and institutional diversity.