TRAVEL TIME RELIABILITY ESTIMATION: USE OF MEDIAN TRAVEL TIME AS MEASURE OF CENTRAL TENDENCY

摘要

This paper proposes use of median travel time compared to mean travel time as the measure of central tendency to calculate Buffer Time Index (BTI), one of the most important measures of travel time reliability. This study examined gamma, exponential, largest extreme value, smallest extreme value, logistic, Weibull, lognormal and log-logistic distributions to determine the best fit to travel time data for two cases: 24 hours and peak hours for a highway with Variable Speed Limit (VSL) conditions. The goodness-of-fit tests indicated that the log-normal distribution best represented the travel time data for peak periods and log-logistic distribution better fitted the travel time data for 24 hours. The log-logistic distribution and log normal distributions were positively skewed and for skewed distributions, median is the best estimate of central tendency. Hence, median travel time is proposed for calculating BTI. The difference in values of BTIs were evaluated when mean and median travel time were used. Travelers must add extra time to their average travel time to ensure on-time arrival. It was found that this extra time increases from 11% to 30% when median travel time is used as the average travel time compared to mean travel time for calculation of BTI.