©Antti Talvitie, Michael Morris, Mark Anderson, 2016
Transportation Research Record (1981) (with M. Morris and M. Anderson).
ABSTRACT
ABSTRACT
The chapter discusses the uncertainties and errors in level-of-service data, in land-use and socioeconomic data, and in behavior data, and their effects on models and forecasts of travel behavior. Evidence is shown to suggest that there are substantial errors in transportation level-of-service data and in forecast land-use and socioeconomic data. Of the service variables, the excess time components and auto costs are especially poorly approximated. There also is uncertainty about what car costs ought to be calculated, and how. It is shown that these decisions affect the choice models significantly. The allocation of land-use and socioeconomic variables to traffic zones is done quite inaccurately, even though the totals for the region may be predicted well. Population forecasts, which may mean household-size forecasts, appear to be subject to a large degree of uncertainty. Finally, the chapter discusses briefly the uncertainties present in travel-behavior interviews. It is claimed that intrasubjective and intersubjective uncertainties are inevitable in any interview and that their effects are many and substantial but presently unknown.
©Antti Talvitie and Youssef Dehghani, 1980
Aggregate Travel Demand Analysis With Disaggregate or Aggregate Travel Demand Models
Transportation Research Forum Proceedings, pp. 583-604, August 1973, R.B. Cross Co., Oxford, Indiana.
Antti Talvitie (aptalvitie(at)gmail.com)
ABSTRACT
THE PURPOSE OF THIS PAPER is to develop a method of aggregation to be used in travel-demand analysis and forecasting. This is an important task because aggregation is always needed in travel demand anal sis. Transportation systems and people need to be aggregated geographically into zones to facilitate demand forecasting aggregation of heterogenous modes of travel is often done to simplify the analysis process; heterogenous people are lumped together for fewer market segments to reduce the size of the problem and so forth.
©Antti Talvitie, Youssef Dehghani, 1979
The paper compares two types of measurements of trip times: those provided by the standard network algorithms are compared with trip-time components observed along the traveler's path from home to work and back. The two types of measurements are found to be different. The root mean square errors of the network measurements with respect to observed values are very large (75-135 percent of the mean value) for the non-line-haul travel time components. The means and the variances of the network measured variables, as a rule, are much smaller than the variances or means of the manually coded observed-travel times. Coefficients estimated by using the two types of data are not numerically similar. Statistical tests show that at least the alternative-specific constants' and the level-of-service variables' coefficients are different in the models developed by using the two types of data. Finally, the effect of substantial errors in level-of-service measurements on travel forecasts is discussed. It is also shown that good (short-run) travel forecasts can be obtained from the network-based models provided that consistent network coding conventions are followed and incremental forecasts are avoided.