Multilevel models to analyze before and after speed data

Abstract

Analyzing before–after speed data is often limited to a standard comparison of various speed parameters. Although a few studies have used a model-based approach, various limitations exist in terms of both data and methodology. The aim of this paper was to examine the applicability of using multilevel models to analyze before–after speed data and to explore the effect of various temporal, geometrical, and traffic characteristics on traffic speed in an urban residential context. Two multilevel models, one with homogeneous and one with heterogeneous within-site variance, were used for analyzing the hourly free-flow speed data. The study used a dataset collected before and after a posted speed limit (PSL) reduction from 50km/h to 40km/h; the reduction was a pilot program in the city of Edmonton, Alberta, Canada. The results demonstrated the appropriateness of using the multilevel model for analyzing speed data. Moreover, the heterogeneous within-site variance model outperformed the homogeneous counterpart in terms of goodness-of-fit and the precision of parameter estimates. The parameter estimations demonstrated intuitive findings with respect to the effect of various factors on mean free-flow speed. In general, the evaluation results showed that the mean free-flow speed was reduced by 4.6km/h in the after period, when the multilevel model with heterogeneous within-site variances was used. The use of the multilevel model with homogeneous within-site variances slightly underestimated the mean free-flow speed reduction. A separate investigation revealed that the mean free-flow speed was reduced by 5.3km/h and 4.0km/h for local and collector roads, respectively.