A pilot study was undertaken to investigate the relationship between the accident rate of a selected section of highway and several measures of the change of speed per unit time for the same section of highway. The change-of- speed data were obtained using a test car floating with the traffic stream under varying traffic conditions. The test car was driven over six functionally different highway test sites employing a total of five different drivers. In investigating the relationship between change of speed per unit of time vs accident rates on different highways, the following hypotheses were evaluated: (1) there exists a unique speed change distribution function with its associated mean and standard deviation for each driver on a given highway, (2) given a highway and a number of vehicles on that facility, the flow characteristics of these vehicles forming the composite traffic stream will yield a speed change distribution function with its mean and standard deviation that will be unique for that facility, (3) the mean and standard deviation for the speed change distribution of the traffic stream will be significantly different for various functional highway classifications, and (4) the mean and standard deviation of a highway facility's speed change distribution function is directly related to the facility's accident rate. For any given driver on a given road, the acceleration noise pattern for that driver is the result of his natural driving characteristics, the roadway geometry, the traffic on the highway, and the nature of the abutting land use. Acceleration noise is thus the aggregate expression of all of these variables. A second traffic parameter, which is mathematically a function of acceleration noise, is the mean of the number of absolute 2-mph speed changes per unit of highway travel time. The Latin squares used in the experimental design permitted the evaluation of the effect of the highway section, the driver, the time of day, the day of the week, and the direction of travel on the mean change of speed. The data were analysed giving the following conclusions: (1) a visual rank order correlation exists between accident rates and the mean of the number of absolute 2-mph speed changes, (2) the variation in the speed change parameter is of such great statistical significance that it overshadows all other sources of variation, (3) the speed change parameter is an extremely sensitive measure of the functional differences in the highways studied, (4) the change of speed data collected for each highway section represent a unique distribution, (5) the mean change of speed parameter used in the analysis is mathematically a function of acceleration noise, and (6) the rank-ordering of the traffic parameter acceleration noise is consistent in all cases with the mean of the number of absolute 2-mph speed changes.
Abstract