Full Bayesian Mixed-Effect Intervention Model for Before–After Speed Data Analysis

Abstract

The analysis of before–after speed data to evaluate the effectiveness of safety interventions is often limited to a non-model-based comparison of speed-related indicators. Moreover, modeling of speed data often does not take into account the nested nature of the data. The objective of the study was to quantify the effect of posted speed limit (PSL) reductions in an urban, residential context. The study employed a mixed-effect intervention model, which could address limitations of existing methods. To model mean free-flow speed and the probability of a speed being below or equal to various thresholds, mixed-effect normal regression and binomial logistic regression models were used. Use of a comprehensive, unique, and disaggregated data set enabled not only the before–after evaluation of the PSL reduction but also the exploration of the effects of various temporal, traffic, and road geometry factors on speed. Results demonstrated the appropriateness of using the mixed-effect model for speed data. Parameter estimations showed that nighttimes; weekends; a high proportion of vans, buses, and trucks; evening peak periods; collector roads; and low hourly traffic volumes were all associated with an increase in the mean free-flow speed and a decrease in the probability of speeds being below or equal to various thresholds. Evaluation results showed the mean free-flow speed was reduced by 3.85 km/h in the after period and that speeds below or equal to 50, 60, 70, and 80 km/h increased by 20.0%, 9.2%, 2.8%, and 0.9%, respectively. All improvements were statistically significant; this finding implied that the PSL reduction was effective at influencing vehicle speed.