5. PROGRAM EVALUATION (continued) 5.3 Speed DataThe ultimate criterion measure for the present study was the speed of vehicles on the test street segments before and after the various countermeasures were implemented. This section details the speed data collection and data management methods, the analytic approach adopted and the results. 5.3.1 Data Collection and ManagementAll speed data for the study were collected using traffic counters. These devices use multiple pneumatic tubes placed on the roadway to record the presence of a vehicle and its speed. The speed data collection was accomplished for the project by the regular traffic count contractors for Phoenix and Peoria as part of the contribution of the two cities to the study effort. All counts were for a consecutive 48-hour period. Multiple counters were placed on each test segment with the number of counters used dependent on both the length of the segment and the features of interest along it. For example, an attempt was made to acquire counts adjacent to newly installed vertical treatments and coincident with the positions of 3-D markings and Tyregrip™ installations. The counters on each segment, although deployed at approximately the same time, did not have their internal timers synchronized. They were put in place, started and subsequently retrieved. Then, during downloading of the data, 48 hours of information was kept and the rest discarded for purposes of these analyses. Counters were located at the same place on each road segment during each measurement period. The counters used by the Peoria contractor recorded an entry for each detected vehicle. This record included the time of the measurement, the vehicle's speed and an estimate of the class or type of vehicle based on the axle length inferred from the data. This vehicle class information was validated as highly accurate by the contractor and the manufacturer of the counter for passenger cars and light trucks. The validity of data for larger vehicle classes such as heavy trucks and buses was suspect. There are many known technical reasons for this pattern of results. For example, if the counter recorded the front wheels of a lead car but missed the rear wheels, the front wheels of a closely trailing vehicle could be erroneously interpreted as the rear wheels of a large truck or bus. Because of this known problem and the fact that there are typically relatively few trucks and buses except delivery vehicles on the test road segments, only cars and light trucks were included in the Peoria data analysis set. In Phoenix , a different type of counter was used that only collected speeds in “bins” or class intervals. The output of these counters was an hourly count of the number of vehicles in the following speed ranges:
It can be seen that these speed class intervals are not of equal size and become tighter in the 30 to 40 mph range. In order to support an analysis of variance (ANOVA) approach to data analysis (see below), the Phoenix data had to be expressed in terms of individual vehicles in the same manner as the Peoria information. This was accomplished by generating a record for each vehicle counted in each interval. The vehicle was assigned a time at the midpoint of the hour in which it was recorded and a speed at the midpoint of the interval in which it was placed by the counter. For example, a vehicle counted in the 26-30 mph (actually 26.00 to 30.99) mph interval from noon to 12:59 p.m. would be assigned a speed of 28.5 mph and a time of 12:30 p.m. This approach disaggregated the class interval data into a set of individual vehicle records for analysis. The effect of this approach is to suppress variance somewhat in the raw data since many speed values (i.e., those not at the midpoint of one of the intervals) simply cannot exist in the data. The data from each city were aggregated into separate databases for analysis. This resulted in files of 435,421 vehicles for Phoenix and 363,143 vehicles for Peoria . These files were imported into the SPSS® Release 11.5 software for analysis8. |
||||