Safety Analyses at Signalized Intersections: Research Strategies, Modeling Techniques, and Significant Factors

Signalized intersections are among the most dangerous locations in a roadway network. The first step towards an effective solution to safety improvement is to identify the primary causes of crash occurrence. Negative Binomial regression provides a common tool for modeling cross-sectional count data like crash frequencies at signalized intersections, which assume the dependent variables are independent. Intersections could be considered as isolated when the distance between them is long. Signalized intersections, especially for those closer ones along a certain corridor, are spatially correlated and will influence each other in many aspects. Generalized Estimating Equations, which is an extension of generalized linear models to correlated data, is applied to account for spatial correlation for the safety analyses at the intersections. Rear-end crashes are the most frequently occurring collision type at signalized intersections. The paper presents a series models developed in the United States for total crashes and rear-end crashes at signalized intersections. Intersection geometric design features, traffic control and operational features, traffic characteristics, land use, and some corridor level factors are collected and their safety effects are tested. The significant factors at the different levels are identified which lead to a better understanding of crash occurrence and eventually to develop efficient safety countermeasures for signalized intersections.