Removing Double-counting from RSI Databases



Removing Double-counting from RSI Databases

Authors

CLAVERING J, Sandman Consultants Ltd, WATSON S, University of Leeds, COLLOP M and HYMAN G, Department of Environment, Transport and the Regions, UK

Description

The Department of the Environment, Transport and the Regions is currently interested in the development of origin-destination matrices, based on roadside interview data, for regional models and for a prospective National Transport Model. Beeanse the costs

Abstract

The Department of the Environment, Transport and the Regions is currently interested in the development of origin-destination matrices, based on roadside interview data, for regional models and for a prospective National Transport Model. Beeanse the costs of roadside interviews are substantial, th e cost-effective re-use of existing data is essential. Use of data from different surveys often involves removing double counting by merging multiple estimates of the same matrix cell, which in turn depends on estimating the accuracy of each cell. A variety of methods have been developed to tackle these problems. The purpose of this study was to consider the statistical theory underlying these different methods and to make recommendations for future practice, based also on practical considerations of handling large datasets. This paper concentrates on the statistical aspects.

In particular two distinct approaches to merging and accuracy were considered, that implemented in the MATVAL suite of matrix-building soRware, and the methods used to create the London Area Transport Survey (LATS) 1991 trip matrices. Some variants of these approaches, those described in the Traffic Appraisal Manual, and those used in the ERICA suite of programs (Davidson 1993) were also considered (see Clavefing and Watson 1997a), but for simplicity this paper is restricted to the MATVAL and LATS methods.

Three stages in the construction of origin-destination matrices are considered (figure 1). In section 2, the sampling assumptions are described which are fundamental to the interpretation and understanding of the interview data. Following this, a factoring stage often takes place, and the issues raised in this are summarised in section 3. One of the most difficult tasks is the edicient merging of multiple estimates of a trip movement and some of the weighting and accuracy problems are outlined in section 4. The conclusions are contained in section 5. References (section 6) provide additional background material to some of the theoretical and practical issues raised.

Publisher

Association for European Transport