LAUSR

Abstract One of the major challenges with vehicle crash frequency studies is how to deal with the unobserved heterogeneity in crash data. While statistical models of crash frequency analysis based on single probability distributions are constantly improving, several researchers discovered that multiple distribution models may better describe crash frequency data and capture more unobserved heterogeneity. Based on the hypothesis that total crash counts occurring at an intersection are affected by unique sets of factors, this research proposes a two-step approach to studying the contributing factors to crashes at intersections in the Mississippi coastal area. In this study, the data of single crash accidents are first clustered into subgroups using a hierarchical clustering method, and then a Random Effects Negative Binomial model is applied to each subgroup with crash counts at an intersection as observations. A model with no data clustering is also estimated to serve as the comparison benchmark. The analysis results show that this two-step approach can reveal more information about crash contributing factors and improve the predictive power and goodness of fit.