RECALIBRATION OF DISAGGREGATE MODE CHOICE MODELS BASED ON ON-BOARD SURVEY DATA

A growing number of transportation planners and modellers use disaggregate logit mode choice models as an element in the demand estimation process. The paper describes a method for recalibration of logit models using widely available on-board survey data. Logit mode choice models are applied on a variety of populations, such as actually surveyed households, "simulated travellers" who are generated by Monte Carlo techniques, "aggregate" travellers who represent whole traffic zones, etc. The estimated choices are aggregated to provide estimates of ridership by mode. In many cases, the data for direct model calibration are not available, and modellers have to rely on published models which do not always fit their particular circumstances. Even in cases where data are available and calibration is performed, sampling and modelling imperfections might result in large errors when aggregate estimates are compared to observed data. On-board surveys are inexpensive, simple to conduct, and are widely available. A properly conducted survey can provide reliable estimates of transit ridership. This paper describes a procedure for adjusting logit model parameters so that the aggregate ridership estimates best approximate ridership estimates from an on-board survey. The procedure addresses the various issues of calibrating a disaggregate model with aggregate data. It takes particular care to account for the highly non-linear nature of the logit model. Through appropriate transformation, the problem is linearized; correlation tables and linear regressions are used in a semi-manual iterative process to adjust the model's parameters. The objective of the process is to maximize the amount of variations which is explained by the model, while approximating the observed transit ridership by submode and district. The paper describes the application of the model in Cleveland, Ohio. (Author/TRRL)

  • Availability:
  • Supplemental Notes:
    • Transportation Analysis and Models. Proceedings of Seminar Q help at the PTRC 10th Summer Annual Meeting, University of Warwick, England.
  • Corporate Authors:

    PTRC Education and Research Services Limited

    110 Strand
    London WC2,   England 
  • Authors:
    • Gur, Y J
  • Conference:
  • Publication Date: 1982

Media Info

  • Features: Figures; References; Tables;
  • Pagination: p. 289-300

Subject/Index Terms

Filing Info

  • Accession Number: 00381639
  • Record Type: Publication
  • Source Agency: Transport Research Laboratory
  • ISBN: 0-86050-113-2
  • Report/Paper Numbers: REPORT
  • Files: ITRD, TRIS
  • Created Date: Mar 30 1984 12:00AM