01333531

Component

One of the most important biases that have been documented in the literature that affects the evaluation of highway safety treatments is the regression-to-the-mean (RTM). Not including it, could over-estimate the effects of the treatment. Although a lot of work has been devoted to the RTM, very few studies have specifically examined issues related to site selection effects for such evaluation. As reported in the literature, site selection effects and RTM are two distinct biases and should be analyzed separately. The primary objective of this paper is to describe how site selection effects can influence the safety effectiveness of treatments. Even when the RTM is accounted for, the selection effects still provide a biased estimate of the intervention. The study objective was accomplished using simulated data. The bias was estimated for the three most common before-after methods: Naïve, using a control group (CG) and for the EB method. Five scenarios were examined: a direct comparison of the methods, sample size, dispersion parameter, the safety effectiveness, and the standard deviation associated with the safety effectiveness. The study results show that among all methods evaluated, the use of a control group (CG) method can eliminate the site selection and the RTM biases, as long as the characteristics of the control group are exactly the same as the one for the treatment group. In practice, this may not be feasible, since the characteristics are seldom the same between the treatment and control groups, especially for crash data. The Naïve method is the one that is the most significantly affected by the selection bias. The EB method reduces the bias, but does not eliminate the total bias. It is therefore important to include the site selection effects when conducting before-after studies.

01329018

11-2868

English

24p

2011

Transportation Research Board 90th Annual Meeting

DVD

Figures (8) ; References (28)

Bias (Statistics); Data collection; Data quality; Highway safety; Risk assessment; Safety; Safety factors; Statistical sampling; Traffic simulation; Validity

Transportation infrastructure

Data and Information Technology; Highways; Safety and Human Factors; I80: Accident Studies; I81: Accident Statistics

Transportation Research Board Annual Meeting 2011 Paper #11-2868

TRIS, TRB

Feb 17 2011 6:19PM