Estimation of origin-destination trip rates in Leicester

The road system in region RA of Leicester has vehicle detectors embedded in many of the network's road links. Vehicle counts from these detectors can provide transportation researchers with a rich source of data. However, for many projects it is necessary for researchers to have an estimate of origi n-to-destination vehicle flow rates. Obtaining such estimates from data obs erved on individual road links is a non-trivial statistical problem, made m ore difficult in the present context by non-negligible measurement errors i n the vehicle counts collected. The paper uses road link traffic count data from April 1994 to estimate the origin-destination flow rates for region R A. A model for the error prone traffic counts is developed, but the resulti ng likelihood is not available in closed form. Nevertheless, it can be smoo thly approximated by using Monte Carlo integration. The approximate likelih ood is combined with prior information from a May 1991 survey in a Bayesian framework. The posterior is explored using the Hastings-Metropolis algorit hm, since its normalizing constant is not available. Preliminary findings s uggest that the data are overdispersed according to the original model. Res ults for a revised model indicate that a degree of overdispersion exists, b ut that the estimates of origin-destination flow rates are quite insensitiv e to the change in model specification.