PURPOSES : The primary purpose of this study is to establish a crash probability model based on a statistical method that explains the relationship between regressor and explanatory variables using both fixed and random effects to control the heterogeneous characteristics of the observed data. In addition, an attempt was made to discover the leading cause of crashes by vehicle type, including passenger car, bus, truck, and trailer. METHODS : The levels of each route and day of the week are grouped using raw expressway crash data for 10 years from 2012 to 2021, and a multilevel mixed-effect logit model is constructed for each vehicle type assuming that the error terms are derived from the hierarchical structure of the group to which they belong. RESULTS : Speeding and obstacles on the road are significant factors that increase the probability of passenger car crashes, and bus crashes have a high rate at toll gates on weekdays. CONCLUSIONS : The multilevel mixed-effect logit model derived in the study has higher accuracy than the general logit model, confirming that mixed-effect analysis is plausible.

The research contributes extending and reviewing of restricted (constrained) and unrestricted (unconstrained) models in GLM(Generalized Linear Models). The paper includes the methodology for finding EMS(Expected Mean Square) and F0 ratio. The results can be applied to the gauge R&R(Reproducibility and Repeatability) in MSA(Measurement System Analysis).