Quantile Regression Analysis of Transit Travel Time Reliability with Automatic Vehicle Location and Farecard Data

Transit agencies increasingly deploy planning strategies to improve service reliability and real-time operational control to mitigate the effects of travel time variability. The design of such strategies can benefit from a better understanding of the underlying causes of travel time variability. Despite a significant body of research on the topic, findings remain influenced by the approach used to analyze the data. Most studies use linear regression to characterize the relationship between travel time reliability and covariates in the context of central tendency. However, in many planning applications, the actual distribution of travel time and how it is affected by various factors is of interest, not just the condition mean. This paper describes a quantile regression approach to analyzing the impacts of the underlying determinants on the distribution of travel times rather than its central tendency, using supply and demand data from automatic vehicle location and farecard systems collected in Brisbane, Australia. Case studies revealed that the quantile regression model provides more indicative information than does the conditional mean regression method. Moreover, most of the coefficients estimated from quantile regression are significantly different from the conditional mean–based regression model in terms of coefficient values, signs, and significance levels. The findings provide information related to the impacts of planning, operational, and environmental factors on speed and its variability. On the basis of this information, transit designers and planners can design targeted strategies to improve travel time reliability effectively and efficiently.