Analyzing factors contributing to real-time train arrival delays using seemingly unrelated regression models

Understanding the impact of various factors on train arrival delays is a prerequisite for effective railway traffic operating control and management. Existing studies analyze delay using single, generic regression equation, restricting their capability in accounting heterogeneous impacts spatiotemporal as travels along its route. The paper proposes set equations conditional location analyzing at stations. We develop seemingly unrelated equation (SURE) model to estimate coefficients simultaneously while considering potential correlations between residuals caused by shared unobserved variables among equations. data from 2017 2020 Sweden are used validate proposed explore effects delays. results confirm necessity developing station-specific models understand explanatory variables. show that significant impacting primarily operations, including dwell times, running operation previous trains upstream calendar, weather, maintenance also Importantly, different management strategies should be targeted stations since these could vary depending where station is.