Fundamental to the present research is the use of hourly traffic volumes in regression models for estimating accident potential on two-lane rural roads. By using data from Ontario, Canada, a simple model form, and a regression package that allows the assumption of a negative binomial error structure, regression models were calibrated for the different combinations of time periods (24 hr, day hours, and night hours) and geometric (roadway and shoulder width) characteristics. It is shown that the effect of day/night conditions is different for single-vehicle and multivehicle accidents. For single-vehicle accidents the accident potential is higher during the night, whereas for multivehicle accidents the opposite is true. This indicates the importance of differentiating between single-vehicle and multivehicle accidents and day/night conditions. The refinement of the regression predictions by the empirical Bayesian (EB) estimation procedure for individual road sections is illustrated. It is shown through a validation exercise that the EB procedure provides better estimates of accident potential than the conventional method only on the basis of the short-term accident count for a section.


  • English

Media Info

  • Features: Figures; References; Tables;
  • Pagination: p. 134-139
  • Monograph Title: Human performance and safety in highway, traffic, and ITS systems
  • Serial:

Subject/Index Terms

Filing Info

  • Accession Number: 00713545
  • Record Type: Publication
  • Files: TRIS, TRB, ATRI
  • Created Date: Nov 13 1995 12:00AM