When treatment is applied to road sections, intersections, drivers, or vehicles that have had a poor accident record in the past, a simple before-and-after comparison of accidents will usually make useless treatments appear effective and overestimate the effect of useful treatments. Because much of what we know about the effect of various safety countermeasures comes from such studies, it is important to demonstrate that this bias can be very large, to show that it can be relatively easily purged from before-and-after comparisons, and to examine whether the method works. These are the three central aims of this paper. To render the statements credible, several data sets are used. The data come from Canada (Ontario), Sweden, the United Kingdom, Israel, and the United States (North Carolina and California) and relate to road sections, intersections, traffic circles, driver violations, and driver accidents. These data sets are used to demonstrate the magnitude of the bias, to illustrate the technique for its elimination, and to examine the success of this debiasing procedure. (Author)
[1]
U Brüde,et al.
Regression-to-mean effect : Some empirical examples concerning accidents at road junctions
,
1982
.
[2]
Ezra Hauer,et al.
Bias-by-selection: Overestimation of the effectiveness of safety countermeasures caused by the process of selection for treatment
,
1980
.
[3]
H. Robbins.
Prediction and estimation for the compound Poisson distribution.
,
1977,
Proceedings of the National Academy of Sciences of the United States of America.
[4]
R D Helliar-Symons,et al.
YELLOW BAR EXPERIMENTAL CARRIAGEWAY MARKINGS - ACCIDENT STUDY
,
1981
.
[5]
C Abbess,et al.
Bayesian Methods Applied to Road Accident Blackspot Studies
,
1983
.