Seemingly Unrelated Regression Model of Car Speeds and Speed Deviations in Freeway Work Zones

Abstract

Current work zone design and traffic control guidance is based heavily on desirable speed-related outcomes, but knowledge concerning actual speed-related outcomes of design and traffic control decisions is limited. The objective of this research is to investigate relationships between speed behavior, roadway and roadside geometrics, and traffic control in work zones. The objective is accomplished through specification and estimation of a seemingly unrelated regression (SUR) model. The dependent variables modeled were 85th-percentile passenger car speed and standard deviation of passenger car speed. Work zone design and traffic control features were investigated as explanatory variables. Speed and infrastructure data used for estimation were collected in Pennsylvania and Texas work zones. The SUR model accounted for contemporaneous correlations of the disturbance terms in the two speed equations. In the equation for 85th-percentile speed, regression parameters were statistically significant for variables representing work zone configuration, type of roadway infrastructure, work zone location, distance traveled from the beginning of the work zone, posted speed limit, vertical alignment, and total paved cross-section width. In the equation for standard deviation of speed, parameters were statistically significant for variables representing distance traveled from the beginning of the work zone, total paved cross-section width, and reduction in posted speed and roadside conditions. Several recommendations for future work are provided, including expansion of the system of equations to include 85th-percentile truck speeds and standard deviation of truck speeds and consideration of possible contemporaneous relationships between speed measures.