Predictive Modeling of COVID-19 Math Learning Loss Using Educational Data Mining and Longitudinal Growth Trajectories

• In Event: AERA Poster Session Wednesday 1:45pm
  In Poster Session: Data-Driven Decision Making in Education SIG Poster Session

Abstract

This study employed educational data mining and longitudinal growth modeling to estimate the impact of COVID-19 disruptions on students’ mathematics achievement. Using STAAR data from a large Texas school system, we trained a predictive model on pre-pandemic cohorts to simulate expected Grade 6 and Grade 7 scores for students who missed testing in 2020. A paired-sample t-test revealed that observed Grade 7 math scores in 2021 were significantly lower than predicted scores, with a large effect size. These findings confirm substantial pandemic-related learning loss. The study demonstrates the utility of predictive modeling in contexts of missing data and offers a scalable framework for assessing academic recovery and informing post-pandemic educational interventions.

Authors

• Zhuoying Wang, University of Texas at Austin

• Shifang Tang, East Texas A&M University

• David Daniel Jimenez, Texas A&M University - Corpus Christi

• Mei Jiang, East Texas A&M University

• Lei Zhang, East Texas A&M University