An investigation to attempt to develop a technique for measuring the driving hazard of one highway relative to another began in 1966. Data for this study were collected with a drivometer-equipped test car operating on six functionally different highways. This paper is an outgrowth of the driving hazard study and was specifically intended to explore any correlation that might exist between the recorded drivometer events and traffic volumes on a given facility. Data were collected from a driving schedule which utilized a Latin square experimental design to minimize bias between drivers, times, and facilities. Eight drivometer events, each recorded on a 0.1-mi basis in one direction on an urban arterial street, were selected as independent variables to be compared with 15-min traffic volumes on an urban arterial street in Raleigh, N.C. these drivometer events included total travel time, change in speed, running time, small steering reversals, large steering reversals, brake applications, accelerator applications, and changes in direction of travel. For the facility under investigation, the mean number of brake applications correlated most closely with traffic volumes. However, none of the simple correlations was high enough to assume complete linear dependence of traffic volume on any single drivometer event. By combining the two most significant events (brake applications and large steering reversals) and using a multiple regression analysis, equations were developed to predict traffic volumes on the investigated facility. Although specific findings are obviously applicable only to the study site, it can be concluded that, by measuring a suitable sample of drivers with a drivometer-equipped vehicle and recording their brake applications and large steering reversals, traffic volumes can be computed with statistical confidence. /Author/.
Abstract