Safety Evaluation of Signalized Restricted Crossing U-Turn Intersections

Chapter 4. Selection of Best Comparison Sites

The quality of the result from the comparison group method relies on the selection of high-quality comparison sites. The project team collected data from four States: Alabama, North Carolina, Ohio, and Texas. To ensure high quality, the project team undertook the following steps during comparison site selection:

1. Identified large intersections near the treatment sites using aerial photographs to serve as possible comparison sites. They identified four potential comparison sites for each RCUT. The project team examined the available time series of aerial photos to ensure there were no major changes to the roadways or surroundings near the potential comparison sites during the study periods.
2. Contacted the respective State agencies to confirm the comparison sites were appropriate in terms of the trends in traffic volumes and the surrounding environment.
3. Obtained crash data for the comparison sites and the RCUTs from the State agencies.
4. Employed the odds ratio test (previously discussed) to determine the appropriate set of comparison sites for each treatment site. The project team tested each potential comparison site separately and all combinations of comparison sites.
5. In cases where odds ratio tests showed that all four potential comparison sites—and all combinations of those four—were unfit for use, computed odds ratios for the comparison sites associated with other treatment sites in the same metropolitan area.

After identifying the best comparison sites and computing CMFs for each individual treatment site, the project team analyzed groups of sites. Analyzing groups of sites boosted sample sizes and generally decreased the SDs of the estimates of the CMFs. The project team analyzed the following:

• Groups of treatment sites from the same State.
• Groups of treatment sites with the longest available before periods (Alabama, North Carolina, and Ohio).
• All sites together.

In all group analyses, the project team used all years of available data at each site. This meant that the group results are somewhat biased in favor of the sites with more data.

A difficulty in analyzing groups of sites was determining the best set of comparison sites. The project team again used odds ratio tests to identify the best set of comparison sites, starting its search with the best comparison sites identified during the individual site analyses.

A group analysis built the sample size and averaged the different sites to overcome the physical or data shortcomings present at each site. The group analyses resulted in the overall signalized RCUT CMF that is considered a predictor of future general signalized RCUT safety performance.

Supplemental Analysis

In addition to the before–after analyses of all reported crashes previously discussed, the project team conducted a before–after analysis of fatal and injury crashes. The sample of fatal and injury crashes was much smaller than the sample of all crashes, which further emphasized the importance of the group analyses.

The project team analyzed other key crash variables at the individual site and group levels, including day/night and crash type, to determine any changes between the before and after periods. The variable analyses help signalized RCUT designers identify any issues with early installations.

After the project team estimated the CMF for signalized RCUT installation, they developed a B/C ratio for the improvement. Participating States provided information on the cost of the RCUT installation at each study site. The project team based the benefit on the estimated crash reduction (if any) and on an estimate of the travel time savings experienced by motorists using the RCUT. They derived the B/C ratio employing similar methods used in CMF research and development.