Modelling road accident blackspots data with the discrete generalized Pareto distribution

Area of Science:

• Traffic Safety and Risk Analysis
• Statistical Modeling
• Road Accident Analysis

Background:

• Road accidents on blackspots pose significant risks.
• Traditional models may not fully capture accident data complexities.
• Accurate modeling is crucial for effective traffic safety interventions.

Purpose of the Study:

• To model road accident and fatality data on Spanish blackspots.
• To evaluate the suitability of discrete generalized Pareto and discrete Lomax distributions.
• To compare these models with the negative binomial distribution.

Main Methods:

• Analysis of basic properties (distribution, survival, mass, quantile, hazard functions, moments) of proposed models.
• Application of parameter estimation methods (μ and (μ+1) frequency, maximum likelihood).
• Goodness-of-fit testing using Chi-square and discrete Kolmogorov-Smirnov tests with bootstrap resampling.

Main Results:

• Discrete generalized Pareto and discrete Lomax distributions were applied to Spanish blackspot data (2003-2007).
• Both models demonstrated utility in describing crash and fatality datasets.
• Probabilistic models showed effectiveness compared to the negative binomial distribution, even with covariates.

Conclusions:

• Discrete generalized Pareto and discrete Lomax distributions are valuable tools for modeling road accident blackspot data.
• These probabilistic approaches offer a robust alternative for analyzing traffic safety events.
• Findings support the use of advanced statistical distributions for better understanding and managing road accident risks.