Bayesian methodology to estimate and update safety performance functions under limited data conditions: a sensitivity...

Area of Science:

• Road safety
• Statistical modeling
• Bayesian inference

Background:

• Decision-makers in road safety often face limited data.
• Maximum Likelihood Estimation (MLE) is unreliable and biased under such conditions.
• Bayesian estimates can be biased with non-informative priors in limited data scenarios.

Purpose of the Study:

• To present a practical Bayesian method for estimating and updating safety performance function (SPF) parameters.
• To combine limited data information with existing SPF parameters from the Highway Safety Manual (HSM).
• To offer a more reliable estimation method when data is scarce.

Main Methods:

• A Bayesian updating approach is proposed.
• The method integrates limited observational data with prior information from the HSM.
• Validation is performed using extensive simulations and sensitivity analysis across 15 models.

Main Results:

• The proposed Bayesian method yields reliable estimates with fewer observations.
• Sensitivity analysis demonstrates the method's robustness across various prior distributions.
• The approach contributes to a unified Bayesian updating process for SPFs.

Conclusions:

• The developed methodology accurately estimates and updates baseline SPFs.
• This approach is a promising tool for road safety analysis, especially under data limitations.
• It enables effective evaluation of road safety countermeasures with limited data.