THE BOOTSTRAPPING APPROACH FOR INFERRING CONFIDENT FREIGHT TRANSPORT MATRICES
F.G. Benitez, University Of Sevilla, L. M. Romero, University of Sevilla, N. Caceres, University Of Sevilla
In this communication we present a methodology, based on the techniques of "bootstrapping", to the robust statistical estimation of freight transport matrices; this allows to generate the confidence intervals of travel between origin-destination pairs defined by each cell of the OD matrix derived from a freight transport survey.
Transport studies require, as a preliminary step, conducting a survey process to a sample of the universe of users of the transportation system. The statistical reliability of the data determines the goodness of the results and conclusions that can be inferred from the analyses and models generated. Due to the high economic costs of the survey capturing stages, collected data are customarily reused as a prior piece of
information, in an either disaggregate or aggregate way. In the first case the statistical reliability is not always guaranteed, this drastically affects the outcomes derived from projections and estimates of prospective scenarios.
In this communication we present a methodology, based on the techniques of "bootstrapping", to the robust statistical estimation of freight transport matrices; this allows to generate the confidence intervals of travel between origin-destination pairs defined by each cell of the OD matrix derived from a freight transport survey. This result is basic to define the dimensions of variability in the cells of said matrices
and subsequent adjustment by updating techniques based on aggregate data (i.e. traffic counts, cordon survey, aggregate link counts, observed paths, etc.).
The techniques of "bootstrapping" originated in the 70's, although widely used during the 90's, have not been fully exploited in the field of freight transport studies.
To address this study a data set from a statistically reliable freight transport study conducted in Spain at the level of multi-province regions has been used. The outcome of this research concentrates in the OD matrices at the interprovincial level derived by the application of the bootstrapping approach. The study shows the usefulness of this approach in terms of mathematical simplicity and accuracy. The results are very stimulating for engineers and practitioners involved in transport research, particularly in demand modelling.
Keywords: origin-destination matrix, freight transport survey, matrix estimation, bootstrapping.
Association for European Transport