Thakali, L., Fu, L., Chen, T., & Usman, T. (2013). A Non-parametric Approach to Road Safety Analysis – Does It Make a Difference?. Proceedings of the 2 cialis price walgreens order cialis for daily use
In road safety research, it has long become a tradition to take a parametric approach to modeling road collisions, which has resulted in a variety of parametric models such as Negative Binomial, Poisson lognormal, zero-inflated Poisson, and random-effect models. While easy to apply and interpret, a parametric approach has several critical limitations due to the modeling requirement of assuming a specific distribution form for each model variable and a fixed functional relationship between each model parameter and the predictors. Violation of these assumptions could lead to biased and/or erroneous inferences on the effect of these predictors on the dependent variable (e.g. collision frequency). This paper introduces a data driven, nonparametric alternative – Kernel regression aiming at answering the question of whether or not it makes any meaningful differences as compared to the traditional parametric method. The proposed approach was applied to model winter road collisions using a database containing hourly observations of collisions, road weather and surface conditions, and traffic counts on a set of highways in Ontario, Canada, over six winter seasons. The results from this approach have clearly shown that significantly nonlinear relationships exist between collision frequency and some condition factors, which were not captured in a prior study using a parametric technique. Furthermore, the new approach also captured some moderating effects of several condition variables, which could have been easily missed in a parametric analysis.