9410612 ANSELIN The available methodology to carry out an "exploratory" analysis of geographic data is limited, especially for the analysis of large data sets that combine information from several points in time with large amounts of geographic detail. In such instances, which are often characterized as data rich but theory poor, the discovery of significant patterns of association (across space, over time, and between multiple variables) is a crucial aspect of the initial phase of scientific inquiry (i.e., the formulation of hypotheses). This situation is increasingly encountered in empirical research as large computerized data bases become available to researchers studying such topics as global change, the spread of contagious diseases, and socio- economic dynamics. The research proposed here will remove these impediments to the exploratory analysis of geographical data by developing a set of new methods to assess association in space- time, with special emphasis placed on categorical variables. The new methods will be developed as an extension of the idea of a general class of Local Indicators of Spatial Association (LISA) developed earlier by the principal investigator. The proposed research consists of a rigorous assessment of the statistical properties of the new methods by means of a series of carefully designed simulation experiments. The new techniques will be disseminated to the research community in the form of a software toolbox for exploratory data analysis, based on an object oriented design. The potential of the new methods for providing new insights in the analysis of large geographical data bases will be illustrated in a case study of the human impacts of global change. This project will provide much needed new statistical tools for exploring relationships throughout space and over time and thus taking better advantage of the large data bases that are being constructed and organized in geographic information systems. The project will result in a set of new software tools that will be made accessible to, and have a significant impact on, the large research and business community who use such data bases.
