Using time-series designs to estimate changes in freeway level of service, despite missing data

Abstract

Until recently, time-series designs have seen little use in evaluating freeway management schemes, but the automatic collection of freeway traffic flow data by permanent detectors now makes such analyses feasible. Since detector failures and other factors may produce data gaps in a series, practical, general models must permit multivariate estimation despite some missing data points. Recent developments in time-series analysis make this possible. Using time-series regression analyses (Harvey and Philips, 1979; Jones 1985), it is possible to detect relatively small average changes in traffic flow characteristics, such as peak hour volume and lane occupancy. These can then be related to the freeway's level of service.