Validation of the influencing factors associated with traffic violations and crashes on freeways of developing countries: A case study of Iran

• Published in: Accident analysis and prevention Vol. 121; pp. 358 - 366
• Main Authors: Hadji Hosseinlou, Mansour, Mahdavi, Alireza, Jabbari Nooghabi, Mehdi
• Format: Journal Article
• Language: English
• Published: England Elsevier Ltd 01.12.2018
• Subjects: Accidents, Traffic - mortality; Accidents, Traffic - statistics & numerical data; Automobile Driving - legislation & jurisprudence; Automobile Driving - statistics & numerical data; Bayes Theorem; Chi-Square Distribution; Chi-square test; Crash; Developing Countries; Freeways; Humans; Iran - epidemiology; Motor Vehicles - classification; Motor Vehicles - statistics & numerical data; Poisson; Poisson Distribution; Risk Assessment; Violation; Zero-truncated Poisson; Violation; Freeways; Crash; Poisson; Chi-square test; Zero-truncated Poisson
• ISSN: 0001-4575; 1879-2057; 1879-2057
• DOI: 10.1016/j.aap.2018.06.009

•The number of crashes per kilometer in Iran’s freeways is almost twice the other rural roads.•This study aims to validate the amount of influence of factors associated with traffic violations and crashes in freeways.•The proposed models are normal Poisson and zero-truncated Poisson.•The zero-truncated Poisson model proved to be the best model.•The results indicated that violations have a significant relationship with crashes. Among the rural roads, freeways have the maximum allowable speed limit. This subject increases the tendency of drivers to use these kinds of roads, and despite its positive effects, it has caused numerous problems. One of them is the increase in the rate of traffic violations and crashes. The amount of crashes per kilometer in Iran’s freeways is almost twice the other rural roads. Hence, finding a solution to this problem is of particular importance. This research intends to validate some of the influencing factors which cause traffic violations and crashes in freeways and determine their amount of influence through statistical models. For this purpose, the authors considered violations and crashes for 36 road segments as dependent variables and other factors as independent ones. Since dependent variables were count, discrete, and non-zero, the proposed models were Poisson and Zero-truncated Poisson. The processing of the models indicated that the amounts of Akaike Information Criterion (AIC) and Bayesian Information Criterion (BIC) indices for the Zero-truncated Poisson model are less than those of the Poisson model and the result of the Pseudo-R2 test for this model is within the acceptable range. Moreover, the result of Chi-square test which shows the proximity of expected and observed amounts was better for Zero-truncated Poisson model. Thus, this model has a considerable advantage against Poisson model. Final models indicated that the average speed has a positive correlation with the number of violations and crashes and as it increases, they increase too. Besides, peripheral landscapes, number of interchanges, number of passing lanes, and exemption from paying toll have an opposite relationship with violations and crashes.