01025584

Component

The identification "hazardous" roadways and presentation of evidence to support a recommendation for treatment are elusive problems, especially for a roadway corridor where the traffic exposure is small. The addition or subtraction of a single fatality event can dramatically inflate or deflate the fatality rate making misclassification a strong possibility. Classical statistical methods for making inferences about fatality rates under this condition are ineffective. By incorporating a traffic safety standard or target, Bayesian data analysis for the calculation of failure probability -the probability of failing to meet the standard- an analytically-based method that reflects the uncertainties associated with identification and the decision to treat are brought into an objective, decision-making process. Simple and hierarchical Bayesian models are presented. Hierarchical models are preferred because they naturally pool information from data that are considered related. Observed or "raw" fatality and traffic exposure data from (1) local corridors are assumed to be samples from (2) corridors with similar roadway and roadside properties. A case study is presented to illustrate the approach.

01020180

06-0700

English

21p

2006

Transportation Research Board 85th Annual Meeting

CD-ROM

Figures (3) ; References (24) ; Tables (4)

Bayes' theorem; Crash causes; Crash rates; Fatalities; Hazard analysis; Mathematical models; Poisson distributions; Statistical analysis; Traffic safety

Data and Information Technology; Design; Highways; Safety and Human Factors; I82: Accidents and Transport Infrastructure

Transportation Research Board Annual Meeting 2006 Paper #06-0700

BTRIS, TRIS, TRB

Mar 3 2006 10:26AM