U.S. Department of Transportation
Federal Highway Administration
1200 New Jersey Avenue, SE
Washington, DC 20590
Federal Highway Administration Research and Technology
Coordinating, Developing, and Delivering Highway Transportation Innovations
|This report is an archived publication and may contain dated technical, contact, and link information|
Publication Number: FHWA-HRT-10-043
Date: September 2010
Effects of Yellow Rectangular Rapid-Flashing Beacons on Yielding at Multilane Uncontrolled Crosswalks
CHAPTER 5. EXPERIMENT 3
Participants in experiment 3 consisted of drivers and pedestrians across 22 sites, with 19 sites in Florida, 2 sites in Illinois, and 1 site in Washington, DC. These sites, along with the ADT and posted speed limit at the crosswalk location, are presented in table 3.
Additional participants consisted of drivers and pedestrians located at two school crosswalks in Illinois, one crosswalk in Washington, DC, and one of the sites in St. Petersburg, FL, equipped with an advance warning rapid-flash device similar to the one in Washington, DC. The first site is located at Hawley Street east of Atwater Drive in Illinois, the second site is located at Midlothian Road south of Kilarny Pass Road in Illinois, the third site is located at Brentwood Road and 13th Street NE in Washington, DC, and the fourth site is located at 1st Avenue N and 61st Street in St. Petersburg, FL (see table 3).
The treatment in this experiment is identical to that of experiment 1. The RRFB system as described previously was employed in this study. Exceptions are found at the third and fourth sites. These locations had a device similar to the previous locations with the exception of being equipped with an advance warning rapid-flash sign. The additional sign was a standard STOP FOR PEDESTRIANS AHEAD sign in Washington, DC, and a standard pedestrian silhouette sign at 1st Avenue in St. Petersburg, FL, equipped with an RRFB system similar to those used in the previous experiments. The advance warning sign in Washington, DC, was placed in the approximate area of the ITE threshold previously discussed. This location was designed so that upon activation of the pedestrian call button, the advance sign would activate immediately. After approximately 1.5 s, the devices located at the crosswalk would then become activated. However, the advance sign in St. Petersburg, FL, was located further away at 368 ft.
This experiment used a before-after design. The baseline was collected for a series of 22 sites. Because these beacons were introduced at different times at each site, it is not likely that the resulting changes were due to any uncontrolled confounding variables such as the level of police enforcement or the occurrence of increased publicity that sometimes follows major pedestrian crashes. After the baseline data were collected, a treatment consisting of either two- or four-beacon RRFB systems was implemented. This treatment was extended in intervals of 7, 30, 60, 90, 180, 270, and 360 days, respectively. Not all sites were yet reporting data to 360 days. The site in Florida equipped with the advance warning sign was evaluated in an alternating treatment design. After a baseline period, the two treatment conditions, the rapid-flash device at the crosswalk sign and the rapid-flash device at the crosswalk sign plus the rapid flash device at the advance warning sign, were alternated in rapid succession (every other crossing).
The general statistical methodology used in this study was based on the general time-series intervention regression modeling approach described in Huitema and McKean and McKnight et al. (See references 8–11.) However, the specific parameters included in the present model differ from those used in the earlier work.
The statistical model used here was developed to conform to the nature of traffic data collected in this study. Because it is well known that compliance with traffic-signal stimulus changes usually occurs rapidly but does not reach an asymptote immediately, the analysis was designed to model this expected change pattern.
Specifically, the change model contained five parameters. The first parameter measured the baseline level, the second measured the change from the baseline to day 7, the third measured the change from day 7 to day 30, the fourth measured the change from day 30 to day 60, and the fifth measured the slope during the remaining time points (days). This fifth parameter measured the general trend after the first month of observations through the final observation month (day 720). An additional parameter was also included to accommodate possible autoregressive patterns in the errors of the model. Because this parameter was of limited interest in this study, it is not described in detail here. The approach used to estimate the parameters of the model is based on a double bootstrap methodology that accommodates both independent and autocorrelated error structures encountered in time-series intervention designs of the type used in behavioral research.(11) Certain variants of this approach have been developed for the analysis of both simple and complex versions of single-case designs.(12)
The five main parameter estimates obtained in the study are shown in table 4. Alpha was set at 0.05 before the data were collected, and any p-value that is less than equal to or 0.05 is statistically significant. P-values are presented to allow the reader to decide whether the evidence is convincing. There is an immediate and large statistically significant level change from the baseline to day 7, a small but statistically significant additional increase from day 7 to day 30, a minor and not statistically significant level decrease at day 60, and a general trend after day 60 that has little slope across the remaining observation days. Hence, the evidence for change is overwhelming, and it is maintained for the duration of the study. There are 144 degrees of freedom for all tests shown in table 4.
The average combined yielding percentage during the baseline of all 19 Florida sites was less than 1.7 percent. Follow-up data were available for all 19 sites at the 7-, 30-, and 60-day periods. The average yielding percentage of all combined sites was 79 percent after 7 days, 86 percent after 30 days, and 82 percent after 60 days. Yielding percentages for the 19 sites at 90, 180, 270, and 365 days were 80, 76, 86, and 83 percent, respectively. The 17 sites that were installed for 2 years showed a yielding compliance of 85 percent 730 days after installation.
Each of the two locations in Illinois has reported data during the baseline and again 7 and 30 days after installation. The first location, Hawley Street east of Atwater Drive, produced 19 percent yielding during the baseline, 71 percent 7 days after installation, and 68 percent 30 days after installation. The second location, Midlothian Road south of Kilarny Pass Road, produced a yielding percentage of 6.6 percent during the baseline. The device was activated 7 days after installation, and yielding compliance increased to 62 percent 30 days after installation. Both of the sites used only two of the rapid-flash devices.
The Washington, DC, location, which was equipped with an advance warning rapid-flash device, was evaluated during baseline conditions and again 7, 30, and 180 days after installation. Baseline yielding compliance at this location was 26 percent. Average yielding compliance increased for 7-, 30-, and 180-day evaluations to 62, 74, and 80 percent, respectively.
The St. Petersburg, FL, site that was equipped with the advance warning device at 1st Avenue North and 61st Street had an average yielding compliance of 8.6 percent during the baseline condition. During activation of the rapid-flash device, average yielding increased to 92 percent only at the crosswalk. The addition of the advance warning device had no effect on yielding, which remained at 92 percent (see table 5).
Data on the distance of yielding drivers were recorded for both of the Illinois sites, the Washington, DC, site, and the St. Petersburg, FL, site at 1st Avenue North and 61st Street that was equipped with the rapid-flash advance warning device. The total combined percentage of drivers yielding at 30 ft or more during the baseline for the two sites in Illinois was 83 percent. The introduction of the treatment device produced increases in the percentage of drivers yielding at 30 ft or more to 94 percent at the Atwater Drive site and 92 percent at the Kilarny Pass Road site. The Washington, DC, site had a baseline percentage of 41 percent for drivers yielding at 30 ft or more. Once the rapid-flash device, including the advance warning sign, was activated 7 days after installation, the percentage increased to 62 percent. Follow-up data collected at days 30 and 180 showed an additional yielding increase at 30 ft or more to 72 and 87 percent, respectively.
The St. Petersburg, FL, site had an average baseline yielding percentage of 50 percent for drivers who yielded at 30 ft or more. No drivers yielded at more than 100 ft during the baseline for this location. During the crosswalk alone condition, the average percentage of those yielding at 30 ft or more was 83 percent. The crosswalk plus advance warning condition saw a slight increase in yielding to 84 percent.