Zone design for environment and health studies using pre-aggregated data

Many environment and health studies employ geographical areas as the units of analysis, either through choice or necessity. The design of these areas can greatly influence any observed spatial relationships or patterns-an effect known as the modifiable areal unit problem. In this paper we identify the phenomena and processes which are typically measured in environment and health studies and present a conceptualisation for their representation as data objects in spatial analysis. We discuss the circumstances under which we find ourselves using areas for representation and outline the application of zone design techniques for the creation of such areas in environment and health studies. An empirical study of the relationship between deprivation and limiting long-term illness in the former county of Avon, UK, is employed to demonstrate the potential usefulness of zone design techniques for creating zones with stable estimates and for exploring the sensitivity of relationships to changes in the zoning system. In particular, we illustrate the inappropriateness of the 1991 Census enumeration district and ward zoning systems for such an analysis and conclude that automatically designed aggregations may be a more appropriate basis for analysis than any pre-existing zoning system. (c) 2004 Elsevier Ltd. All rights reserved.