Functional Approximations to Alternative-Specific Constants in Time-Period Chioce-Modelling

This paper discussed issues researchers are faced with in the case of models using a high number of constants that have a high level of interdependency. The example used in the analysis is that of time-period choice, where constants are used to capture period-specific unobservable along three dimensions; outbound, return and duration. The potentially high number of constants to be estimated in a fully specified time-period choice model can lead to issued of identification, high computational cost, and problems with interpretation. Because of the heightened correlation between adjacent time-periods, some form of interdependency is clearly expected. This, in conjunction with the above discussed problems with a fully-specified model, was the motivation for replacing the full set of constants by a functional approximation. This paper proposes several alternative approaches based on parametric as well as non-parametric functional approximations. The analysis has shown that the different approaches all have advantages as well as disadvantages, but that important gains in identification can in general be obtained with the use of a functional approximation.

Language

  • English

Media Info

  • Media Type: Print
  • Features: Figures; References; Tables;
  • Pagination: pp 545-564
  • Monograph Title: Transportation and Traffic Theory. Flow, Dynamics and Human Interaction. 16th International Symposium on Transportation and Traffic Theory

Subject/Index Terms

Filing Info

  • Accession Number: 01002774
  • Record Type: Publication
  • ISBN: 0080446809
  • Files: TRIS
  • Created Date: Aug 5 2005 9:40AM