Spatial Analysis

Analysis can be defined as spatial if the locations of the objects in space matter by affecting the results of the analysis. A simple regression analysis of the relationship between the number of golf courses in a country and its gross national product would not be spatial, for example, because countries could be relocated without affecting the results; the analysis would not require the locations of the countries. To a geographer, spatial analysis almost always implies that the space of interest is geographic space—the surface and near surface of the earth—but methods of spatial analysis might also, in principle, be applied to phenomena distributed within the space of the human brain, along the linear space of the human genome, or on the surface of another planet. In all such cases, methods of spatial analysis might be used to reveal what is otherwise hidden or only partially visible to the user in the form of patterns, anomalies, clusters, or correlations. For example, a classic problem in spatial analysis is to determine whether or not clusters exist in the point patterns created by instances of disease and the implications of those clusters for epidemiological processes.

Many methods of spatial analysis have been devised to answer a wide range of questions. One way in which to organize them is by the type of data they address. Thus, there are methods to analyze the following:

• Patterns of points such as records of the instances of a disease, the locations of crimes, or the locations of nesting birds
• Patterns of lines such as the tracks of hurricanes or migrating birds
• Patterns of areas such as voting districts, sales districts, or patches of habitat
• Surfaces of continuous variation such as images collected from Earth-observing satellites or the earth's topographic surface
• Patterns of interactions between places such as those created by migration, airline travel, or e-mail messages

Today, most methods of spatial analysis are available in the form of computer software. In many cases, software incorporates functions to create, edit, and display data as well as functions for analysis, and such packages are termed geographic information systems (GIS) if they include facilities for handling data referring to locations on the surface of the earth. In other cases, routines for spatial analysis may have been added to standard statistical packages, such as S and SAS, or to mathematical packages, such as MatLab. However, these generally lack the elaborate supporting functionality for dealing with spatial data.

Several distinct ways of organizing the numerous methods of spatial analysis have been proposed to[Page 447] make it easier for investigators to find appropriate methods and appropriate tools. The five data types listed earlier form the foundation for one method, and several textbooks have been organized in this way, discussing in turn methods available for analyzing each major data type. The theory of geographic information suggests, however, that a more fundamental distinction between spatial data types exists and that the point/line/area/surface/interaction classification masks this distinction and so can be confusing. Geographic information scientists first distinguish between continuous fields and discrete objects as two alternate and mutually exclusive conceptualizations of the geographic world.

...