![Sample our Built Environment journals, sign in here to start your access, Latest two full volumes FREE to you for 14 days]({"type":"image" , "src" : "/sda/5926/TOC-Subject-Sample-2021-Built-Environment.jpg"})
ABSTRACT
The article provides an analysis of severe crashes (fatal and injury accidents only) that occurred in 260 Italian road tunnels on the basis of random-parameters regression models. In particular, factors influencing “crash frequency” over a 4-year monitoring period are investigated by means of a random-parameters negative binomial model (RPNB). This approach reveals that some regression model parameters should be considered as random variables, thus contributing to gaining new insights into the way the corresponding covariates influence accident frequency. The presence of a year effect on the number of accidents was also investigated. To this aim, random-parameters models with random-effects were considered to account for temporal correlation in the data collected from the same road tunnel over successive time periods. In particular, a negative binomial model in which the intercept and regression parameters are allowed to vary randomly was developed. It was found that in the present case study the unobserved heterogeneity is small, so it is sufficient just using the random-intercept model. This study also explores the use of crash rate instead of crash frequency as dependent variable of the regression model, to understand more especially the effect of this choice on the significance of the covariates. A random-parameters tobit regression model showed that the significance of some independent variables is in contrast with the results found upon the RPNB model based on crash frequency.