ASSESSING THE SAFETY EFFECT OF TREATMENT USING DATA FROM A NUMBER OF SITES

It is important to find a method of combining information from a number of treated sites and controls, such that comparisons of the traffic accident frequency before treatment with that after treatment will lead to correct estimates of the effectiveness of the treatment. This paper is concerned with methods which are appropriate for the usual cases where data are available before and after periods at the treated site, and for the same periods at the controls; the periods will be typically of two or three years' duration. All of the methods described here can be found in the statistical literature. None are original, but they do not appear to be well known to road safety researchers. The next section of the paper reviews the methods for drawing conclusions from a single site; the difficulty of obtaining significant results and precise estimates is illustrated by a simple example. The Yule-Simpson paradox is described and illustrated in Section 3. Methods based on the non-central hypergeometric distribution are described in Section 4, which is the main section of the paper. The methods are applied to two sets of data. The first is a new analysis of the data used in a research project concerned with accident migration. The second set of data consists of before and after accident frequencies for all mini-roundabouts which have been introduced in a large English county. The final section reviews the results of the methods, and considers briefly any limitations in their use in practice.

• Availability:
  • Find a library where document is available. Order URL: http://worldcat.org/issn/1102626X
• Corporate Authors:

  PTRC Education and Research Services Limited

  Glenthorne House, Hammersmith Grove
  London W6OL9,   England 
• Authors:
  • JARRETT, D
• Conference:
  • Traffic Safety on Two Continents
  • Location: Malmo, Sweden
  • Date: 1999-9-20 to 1999-9-22
• Publication Date: 1998

Language

• English

Media Info

• Features: Figures; Tables;
• Pagination: p. 145-162
• Serial:
  • Publication of: Swedish National Road and Transport Research Institute
  • Publisher: Swedish National Road and Transport Research Institute (VTI)

Subject/Index Terms

• TRT Terms: Crash analysis; Crash data; Crash investigation; Crash rates; Statistical analysis; Traffic crashes
• Subject Areas: Data and Information Technology; Highways; Safety and Human Factors; I81: Accident Statistics;

Filing Info

• Accession Number: 00766125
• Record Type: Publication
• Report/Paper Numbers: VTI konferens 9A part 6
• Files: TRIS, ATRI
• Created Date: Jul 30 1999 12:00AM