Frailty Models for the Estimation of Spatiotemporally Maximum Congested Impact Information on Freeway Accidents

The objective of this paper is to develop models for the estimation of the temporal and spatial extent of congestion impact caused by accidents. Although there have been various approaches based on the deterministic queuing diagrams and kinematic wave (or shockwave) theory, only a few studies have been able to estimate the spatiotemporal congested region based on field data, such as ubiquitous loop detector data. Accordingly, this paper applies a previously developed procedure to capture the spatiotemporal accident impacts based on binary integer programming (BIP). The procedure provides a foundation for models of the following: 1) maximum spatial distance to the end of the congestion region affected by each accident and 2) maximum time affected by congestion resulting from each accident. Based on these models, the objective of this paper is to estimate two statistical models for providing maximum congested distance and time information due to freeway accidents. Since various observations from BIP were censored with respect to time and space, survival analysis - specifically, frailty models to account for unobserved heterogeneity - is applied to identify factors critical to spatiotemporal congestion impacts of freeway accidents.