This paper presents some statistical methods and accuracy measures that can assist model development, and gives some illustrative case studies. Errors can be subdivided into: (1) sampling errors; (2) measurement errors; (3) variability; and (4) inherent model errors. Sampling methods are widely used, and include hypergeometric sampling and the jackknife method of estimating sampling error. Four methods are outlined for taking account of inherent error when fitting models. Uses of accuracy measures in model development include: (1) appraisal of old data; (2) survey design; (3) model development from survey data; (4) choice of zone site; (5) the development of infilling and other accuracy models; (6) the specification of accuracy to computer programs; and (7) validation. Data accuracy concepts and methods are illustrated for: (1) an origin destination survey design; (2) guidance in choice of factors; and (3) specification of accuracies for an integrated transport model. There is now enough statistical theory and practical experience to develop a definitive manual of guidance for assessing errors in data as an aid to model development. Such a manual would give practitioners a much-needed sound theoretical framework for handling their problems.
Samenvatting