The representation of a response surface by categorical enumeration of values of the dependent variable has become popular in transportation planning models as an alternative to the more generally known least squares regression method. This popularity is due mainly to the apparently increased freedom given to the analyst since it is no longer necessary to specify, a priori, the algebraic form of the response surface nor to be overly concerned about the effects of heteroscedasticity and multicolinearity on estimated values. Against this must be set the fact that an algebraically defined response surface is better if any but trivial extrapolations are to be made. Initially categories were defined arbitrarily but recently increasing attention has been given to finding acceptable solutions to the problem of category definition. This paper describes one such solution based on minimising the variance of the forecast mean value of the dependent variable. One of the main differences between this and other methods is that it is the forecast values of the independent variables that are used rather than those derived from survey data. This has the effect of emphasising the uncertainty attached to forecasts which is seen as being a thoroughly desirable attribute of the method. /Author/TRRL/

  • Corporate Authors:

    Polytechnic of Central London

    School of Environment, 309 Regent Street
    London W1R 8AL,   England 
  • Authors:
    • Jessop, A
  • Publication Date: 1977-2


  • English

Media Info

  • Features: Figures; References; Tables;
  • Pagination: 32 p.

Subject/Index Terms

Filing Info

  • Accession Number: 00172564
  • Record Type: Publication
  • Source Agency: Transport and Road Research Laboratory (TRRL)
  • Report/Paper Numbers: Monograph
  • Files: ITRD, TRIS
  • Created Date: Jun 28 1978 12:00AM