One of the current methods used to overcome "regression-to-mean" problems in safety studies is to employ a combination of accident histories and accident prediction model estimates for estimating future safety. Besides applications in before-and-after studies, there is also a wide range of applications of accident prediction models (e.g., accident surveillance, network simulation and optimization studies); and they are the focus of this paper. The prediction model used in this paper is a three-level prediction procedure being planned for implementation by the California Department of Transportation (CALTRANS). This staged procedure also allows different applications to be made for a wide variety of projects with different input and output requirements. It is shown that the three-level prediction procedure provides a very detailed, comprehensive, and yet flexible framework for safety evaluations of highway intersections. Meanwhile, one can also appreciate from the discussion that great care should be taken in using those estimates for different purposes. It is also apparent that as accident prediction models are becoming more sophisticated and important to safety studies, a close link should be developed between people who are developing those models and those who are applying them in practice, so that maximum benefits can be obtained.

Media Info

  • Features: Figures; References;
  • Pagination: p. 20-30
  • Monograph Title: Application and management of accident data
  • Serial:

Subject/Index Terms

Filing Info

  • Accession Number: 00495029
  • Record Type: Publication
  • ISBN: 0309049741
  • Files: TRIS, TRB, ATRI
  • Created Date: Jun 30 1990 12:00AM