Quantifying Uncertainties in a National Forecasting Model
ridership - forecasting, ridership - forecasting, ridership - demand, economics - value of time
Value of time, Uncertainty, Travel models (Travel demand), Travel demand, Statistics, Scenarios, Sampling, Projections, Origin and destination, O&D, Mathematical models, Forecasting, Bootstrap technique
This paper seeks to quantify uncertainties that arise from the fact that travel demand models are estimated on a sample of the population rather than the entire population. Forecasting systems can be quite complex, and may contain procedures that do not easily permit analytically derived statistical measures of uncertainty. In this paper, the possibilities of using computer-intensive numerical methods to compute statistical measures for very complex systems, without being bound to an analytical approach, are explored. Here, the bootstrap method is used to obtain statistical measures of outputs produced by the forecasting system SAMPERS. The SAMPERS system is used by Swedish transport authorities. The bootstrap method is briefly described as well as the procedure of applying bootstrap on the SAMPERS system. Numerical results are presented for selected forecast results at different levels such as total traffic demand, origin-destination demand, train line demand and the demand on specific links. The uncertainty related to the value of time estimate also is analyzed. The study demonstrates that the bootstrap method works and is simple to use. However, large bootstrap sample sizes are required for calculating confidence intervals.
Hugosson, Muriel, (2005). Quantifying Uncertainties in a National Forecasting Model. Transportation Research Part A: Policy and Practice, Volume 39, Issue 6, pp 531-547.
Transportation Research Part A Home Page: http://www.sciencedirect.com/science/journal/09658564