Assessment of Methodological Alternatives for Modeling the Spatiotemporal Crossing Compliance of Pedestrians at Signalized Midblock Crosswalks

Abstract

At the midblock crosswalks with either Pedestrian Hybrid Beacons (PHBs) or Traffic Control Signals (TCSs), pedestrians’ crossing compliance can be considered in terms of the space (crossing location with respect to stripes) and time (waiting for a WALK signal). In any crossing incident, two combinations of a scenario are performed jointly; pedestrians can cross within stripes when WALK signal is either active or inactive, also they can cross outside the stripes when WALK signal is either active or inactive. This study presents the assessment of alternative methodologies for modeling the spatiotemporal crossing compliance of pedestrians. It uses data collected from five signalized crosswalks located along four major arterials in Las Vegas, Nevada. Three models, multinomial logit, ordered logit, and logistic regression (LR), are proposed and evaluated. Bayesian information criterion (BIC), Akaike information criterion (AIC), and misclassification error are the three performance measures used to compare the models. Based on these performance measures, the logistic regression outperformed the other two, as it had low AIC and BIC, as well as low misclassification error. This model was then used to evaluate the factors associated with the pedestrians’ spatiotemporal crossing compliance. The logistic regression results revealed that the active WALK sign and the crossing incidences involve female(s) only are positively associated with pedestrians’ spatiotemporal crossing compliance. On the other hand, the optional one/two cross stages, pedestrian wait time, children and teens, as well as people who cross while riding a bike are negatively associated with spatiotemporal crossing compliance.