TESTING DISAGGREGATE TRAVEL DEMAND MODELS BY COMPARING PREDICTED AND OBSERVED MARKET SHARES

An intuitively appealing and popular method for testing a disaggregate choice model of travel demand, such as a logit model, consists of comparing the model's predictions of the market shares of travel alternatives in population groups with observations of these shares. Excessively large differences between the predicted and observed shares indicate that the model being tested is incorrect. In current practice, the decision whether differences between predictions and observations are large is made judgmentally, thereby raising the possibility that a correct or approximately correct model will be rejected because of the effects of random sampling errors. A statistical test is described that enables one to distinguish between the effects of random sampling errors and those of true model errors when predicted and observed market shares are compared. Five easily programmable steps for implementing the test are given, and commercially available software that can help with the computations is identified. A numerical example of the application of the test is presented, and the role of the test in practical model development is discussed

Media Info

  • Media Type: Print
  • Features: References; Tables;
  • Pagination: pp 1-7
  • Monograph Title: TRAVEL MODEL DEVELOPMENT AND OPERATIONS RESEARCH METHODS
  • Serial:

Subject/Index Terms

Filing Info

  • Accession Number: 00395396
  • Record Type: Publication
  • ISBN: 0309037581
  • Files: TRIS, TRB
  • Created Date: May 31 1985 12:00AM