A Generalized Event Count Model for Crash Data Analysis

The investigation of relationships between traffic accidents and relevant factors is important in traffic safety management. Various methods have been developed for the modeling of crash data. In real world scenarios, crash data often display the characteristics of under-, over- or Poisson dispersion. The commonly used models (such as the Poisson and the NB [negative binomial] regression models) have associated limitations to deal with various degrees of dispersion. In light of this, a generalized event count (GEC) model was proposed in this study. This method can be generally used without considering the degrees of dispersion to simplify the process of crash data analysis. This model was applied to case studies using data from highways in Idaho. The results from the GEC model were compared with those from the Poisson regression and the Negative binomial regression models. The cases studies show that the proposed model has good performance for crash data with various degrees of dispersion.