HOW MUCH DATA SHOULD WE COLLECT?

An important problem in transport planning practice is the specification of sample sizes for data collection. To date this appears to have been done by involing various rules of thumb or rather old specifications based on crudely applied 95 per cent significance levels. This paper argues that such rules are not good enough since they ignore the requirements of the decision-making procedure to be used and the inherently probabilistic nature of forecasts. It is therefore suggested that current procedures should be rearranged so that the problem, possible solutions and forecasting and evaluation models are fully specified before any decisions about data collection are taken. Since a Bayesian approach to data analysis is used full scope is given for the incorporation of subjective estimates into the forecasts. (Author/TRRL)

  • Availability:
  • Corporate Authors:

    Whitehall Press Limited

    Headington Hill Hall
    Oxford 0X3 0BW,   England 
  • Authors:
    • Jessop, A
  • Publication Date: 1980-11

Media Info

  • Features: Figures; References;
  • Pagination: p. 22-27
  • Serial:
    • Highway Engineer
    • Volume: 27
    • Issue Number: 11
    • Publisher: Whitehall Press Limited
    • ISSN: 0306-6452

Subject/Index Terms

Filing Info

  • Accession Number: 00330769
  • Record Type: Publication
  • Source Agency: Transport Research Laboratory
  • Files: ITRD, TRIS
  • Created Date: Aug 15 1981 12:00AM