Exploring influential factors and endogeneity of traffic flow of different lanes on urban freeways using Bayesian multivariate spatial models

The traffic flow pertinent modelling is essential for distinct strategy formulations. However, the present literature illustrated some needed improvements for such models, especially those related to lane-specific flow. Given such context, this study aims to bridge the research gap by generating regression models to investigate influential factors for traffic flow based on the data collected from one multilane freeway. With the Full Bayesian specification, the hierarchical models were built by accounting for three types of random effects: the structured and unstructured spatial effects, and the one addressing the multivariate heterogeneity across multiple lanes. The endogenous relationship of traffic flow of adjacent lanes was also explored by utilizing the capability of multivariate correlation structure for simultaneous estimation of lane flow. The model estimates revealed the presence of endogeneity with statistical significance for the flow of neighbouring lanes for both directions of travel. The impact of flow was not only limited to the adjacent lanes but also to non-adjacent lanes. The multivariate specification also confirmed interdependency for lane flows. Compared to conventional approaches, the more accurate model estimation in the study indicates the advantage of incorporating the various correlation structures in the models.