Pedestrian volume data are needed for a variety of purposes - on the demand side, for examining trends and planning facilities, and on the supply side for evaluating safety and level of service, to name a few. However, lack of flexible analytical tools and data gathering costs limit authorities' ability to examine pedestrian flows and provide the required levels of service. Efforts to develop expansion models and demand models to minimise manual data gathering are also hindered by the lack of information about variation in flow over time, as well as uncertainty in the models. A statistical approach based on bayesian theory is discussed that can combine available data, analysts' experience (subjective judgement), and short-period (sample) counts, to estimate the expected (mean) flow at a given site at a given time that will serve as input for a particular expansion model. Compared with sample means, bayesian estimates are much closer to the true mean and, hence, expanded values would be less uncertain. Moreover, the technique could be used to update previous estimates by combining them with newly performed short-term counts. This paper appears in Transportation Research Record No. 1281, Human Factors and Safety Research Related to Highway Design and Operation 1990.
Abstract