The influence of time intervals between test attempts on student performance – an ML-focused analysis of synthetic data

Abstract

The practice of providing students with multiple attempts on assessments is common in many educational settings. However, the optimal time interval between retake attempts and its impact on student learning outcomes remains relatively unexplored. Existing research has shown varying results on the influence of the time interval, with some studies indicating longer intervals leading to increasing performance while others suggest no significant effect. This study analyzes synthetic data generated based on an undergraduate computing course at George Mason University, consisting of two formative tests and one summative test from the second week of classes. A statistical analysis reveals a larger mean time interval in hours between attempts for formative assessments (7.31 and 7.05) compared to summative assessments (4.52) but shows no consistent trend between score gain and time interval. A Random Forest Regression machine learning model was also employed to predict score gain based on time interval. Using data from over 900 unique students, the model generated high mean squared error values along with R-squared values of -0.02399, -0.04447, and -0.07393, indicating very little variance in score gain based on time interval. Other factors, like study methods, should be analyzed alongside time intervals in future research on optimizing academic assessment.