A sampling scheme for quantifying and benchmarking on time performance of urban bus transit

Abstract

In this paper we use large-scale mass transit data to improve the journey time performance measurement of urban bus systems. For low frequency bus services, the application of ontime performance (OTP) metrics, particularly the location and number of stops to sample, varies greatly across operators, which can lead to biased estimates. In this paper, we aim to address sampling disparity, and propose a new statistically robust sampling scheme to ensure representative, non-biased, and comparable measurement of route level on time performance (OTP) which enables the inclusion of operators with sparse data in benchmarking activities. We use automated vehicle location data and analyse 59 unique low frequency route-direction data sets from 9 international bus operators. Across all route-direction data sets, an average of 28% of stops are required to be sampled under the defined sampling scheme to achieve a 95% confidence level of sampling the route-level mean within a ± 5% range of accuracy. By including data on additional routes, the analysis can be further refined and generalised to enable progressive improvement of inter-operator comparability and benchmarking.