MODELS FOR ACCIDENTS AND INJURY CONSEQUENCES TREATED WITH CROSS-VALIDATION AND BOOTSTRAP. REPORT SUMMARY

The negative binomial distribution has been shown to be a practical tool when describing the number of accidents in road junctions of similar types. In this case the distribution is based on the assumption that the number of accidents in a junction follows a Poisson distribution with a specific expected value for the junction, and the expected values for the total population belong to a gamma distribution. Having made an observation (Xobs) of the number of accidents for one of the road junctions, the expected value of the next observation is obtained as E(X) + a(Xobs-E(X)), where E(X) is the expected number of accidents, taken over all the road junctions in the population. The result is valid for road sections as well. The injury consequence has also been treated from this point of view. The parameter "a" can now be estimated by using cross-validation. The variance of this estimate can in turn by estimated with bootstrap technique.

• Availability:
  • Find a library where document is available. Order URL: http://worldcat.org/issn/11015179
• Corporate Authors:

  Swedish National Road and Traffic Research Institute

  Drottning Kristinas Vaeg 25
  S-11428 Stockholm,   Sweden 
• Authors:
  • Junghard, O
• Publication Date: 1989

Media Info

• Pagination: p. 21
• Serial:
  • Nordic Road and Transport Research
  • Volume: 1
  • Issue Number: 1
  • Publisher: Swedish National Road and Transport Research Institute (VTI)
  • ISSN: 1101-5179
  • Serial URL: http://www.vti.se/nordic

Subject/Index Terms

• TRT Terms: Crashes; Injuries; Intersections; Mathematical models; Poisson distributions
• Uncontrolled Terms: Bootstrap technique
• Old TRIS Terms: Cross-validation; Poissons distribution
• Subject Areas: Highways; Safety and Human Factors;

Filing Info

• Accession Number: 00485468
• Record Type: Publication
• Files: TRIS
• Created Date: Jul 31 1989 12:00AM