Geographic visualization software for enumerated data requires careful design if it is to fulfill its potential. This dissertation presents important problems at multiple levels of software design, and presents an approach to both design and integration of multiple EDA/ESDA methods. These design problems range from the design of coordination techniques applicable to diverse problem areas, to design of geographic visualization components for highly multivariate data spaces, to the design of what visual and numerical aspects of the data being represented should be coordinated across components for the creation of an integrated toolkit. This dissertation also presents novel approaches to these problems, and demonstrates working solutions using these approaches, at each level. Coordination among independent components (and the implementation level) and among visualization methods and tools (at the conceptual/operational level) is a major focus of the work. At the level of generic coordination strategies, the use of Java language mechanisms in combination with a standard software structure, leads to an approach that is robust and flexible. At the level of design of individual components, I introduce design ideas from Geovisualization, InfoVis, and EDA, which are implemented to create components appropriate for highly multivariate geospatial data visualization. Finally, the coordination strategies are integrated with the geographic visualization components, to form the GeoViz Toolkit, an analysis environment that is shown to provide novel insights into an important epidemiological data set.
展开▼