Research Papers (2009 – 2013)

Abstract

In road safety research, parametric approaches such as Poisson, Negative Binomial, Poisson lognormal, Zero-inflated Poisson, and random-effect models are commonly used in collision modeling. While easy to apply and interpret, parametric approaches have several critical limitations due to the modeling requirement of assuming a specific probability distribution form for each model variable (e.g. collision frequency) and a pre-specified functional relationship between each model parameter and the predictors. Such assumptions, if violated, could lead to a biased and/or erroneous inference on the effect of these predictors on the dependent variable. This paper introduces a data-driven, nonparametric alternative Kernel regression, which aims to circumvent the need for the aforementioned assumptions. This paper then compares the parametric and nonparametric approaches as they apply to modeling winter road collisions by using a sample dataset consisted of hourly observations of collisions, road weather and surface conditions, and traffic counts from highways in Ontario, Canada, over six winter seasons. The nonparametric approach demonstrates novel and significant nonlinear relationships between collision frequency and some condition factors. It further demonstrates some moderating effects of condition variables that were not captured by parametric techniques.

Lalita Thakali, Liping Fu, Chen Tao