Methodological development for selection of significant predictors explaining fatal road accidents.

Area of Science:

• Road safety research
• Transport policy analysis
• Statistical modeling

Background:

• Accurate identification of road accident causes is vital for effective transport policy.
• Model selection for road safety analysis remains a complex challenge.
• Previous methods have limitations in simultaneously addressing variable and model selection.

Purpose of the Study:

• To develop a novel methodology for model selection in road safety research.
• To identify the most relevant explanatory variables influencing road accident occurrence.
• To provide a robust framework for future transport policy decision-making.

Main Methods:

• A hybrid approach combining neural network design and statistical methods for variable selection (TIM method).
• Modeling the error structure using an autoregressive process.
• Parameter estimation via Markov Chain Monte Carlo (MCMC) with non-informative priors.

Main Results:

• The proposed methodology successfully identified key variables influencing road accident reduction.
• The selected variables are directly relevant to current road safety policy measures.
• Application to Spanish fatal accident data (2000-2011) demonstrated the method's effectiveness.

Conclusions:

• The developed methodology offers a robust approach to model and variable selection in road safety.
• The identified factors provide valuable insights for enhancing future road safety policies.
• This research contributes to a better understanding of road accident determinants and policy impacts.