Modelling trends in road accident frequency- Bayesian inference for rates with uncertain exposure

  • Authors:
  • Louise K. Lloyd;Jonathan J. Forster

  • Affiliations:
  • -;-

  • Venue:
  • Computational Statistics & Data Analysis
  • Year:
  • 2014

Quantified Score

Hi-index 0.03

Visualization

Abstract

Traffic flow data are primarily used to monitor road use and to compute road accident rates in Great Britain. The main traffic flow data used for these purposes measure annual traffic flow in vehicle kilometres, however this dataset is limited in its disaggregation. In particular, it is not possible to quantify traffic flow by different types of cars using just these flow data. Two additional sources of data are introduced (the number of cars registered each year and an induced exposure dataset containing information about different road use by different car types) and a model combines the three datasets in order to produce a disaggregation of traffic flow by car type and road type on the commonly used annual traffic data. MCMC algorithms are used to simulate from the posterior distributions and produce estimates of the traffic by three road types and six car types across 12 years. These flow estimates are then used in models for accident rates.