Applying quantile regression for modeling equivalent property damage only crashes to identify accident blackspots

Area of Science:

• Transportation Engineering
• Traffic Safety Analysis
• Statistical Modeling

Background:

• Current hot spot identification (HSID) methods struggle with underreported minor crashes and skewed data.
• Inefficient HSID leads to misallocated funds and poor risk management.
• Existing safety performance functions often fail with zero-inflated crash data.

Purpose of the Study:

• To propose an improved HSID methodology addressing underreporting and crash severity.
• To develop a technique that accurately identifies high-risk road segments.
• To overcome limitations of traditional methods in handling skewed crash data.

Main Methods:

• Incorporating property damage only (PDO) crash equivalency to account for severity and underreporting.
• Utilizing non-parametric quantile regression to model non-count and skewed crash data.
• Comparing the proposed method with the traditional Empirical Bayes (EB) method using negative binomial regression.

Main Results:

• The proposed method effectively identifies high-risk sites reflecting true societal safety costs.
• It reduces the impact of under-reported property damage only (PDO) and minor injury crashes.
• Quantile regression overcomes limitations of traditional models with zero-inflated and right-skewed datasets.

Conclusions:

• The combination of PDO equivalency and quantile regression offers a more accurate HSID approach.
• This methodology provides a better representation of road safety risks.
• The findings support improved investment decisions and risk management in transportation safety.