Several estimation techniques of route-level trip tables are reviewed and tested. Current industry practice of simple expansion of an on-board survey by total boardings is compared with expansion by the iterative proportional fitting (IPF), constrained generalised least-squares, and constrained maximum-likelihood methods. An intervening-opportunity model, which does not use the on-board survey, is also tested. The more complex methods achieve better accuracy and reduced bias by combining the survey data with ride-check data. An empirical case study demonstrates that under the assumption of error-free ride-check data the IPF technique is preferred because of its computational ease without loss of accuracy. The IPF method should enable transit operators to obtain much more accurate trip tables and more reliable on-board survey results for a small additional computational cost. (A)
Abstract