Recent Advances in Software for Spatial Analysis in the Social Sciences
Rey, Sergio J., Anselin, Luc, Geographical Analysis
During the past decade, numerous research projects have focused on the development and implementation of spatial statistical analysis software. This issue brings together a collection of articles representing recent developments in spatial analysis software for the social sciences. The articles were selected from those presented at two recent conferences on spatial information science. The first was a specialist meeting on "Spatial Software Tools Development," sponsored by the U.S. Center for Spatial Integrated Social Science and held in May 2002 in Santa Barbara, California. (1) The second gathering was a symposium on "Spatial Information Science for Human and Social Science," held at the University of Tokyo in January 2004.
The two meetings covered a wide range of issues from tool development and design to user communities and applications. They brought together a productive mix of software developers, empirical researchers, and private and public sector entities to take stock of the state of the science regarding tools for spatial data analysis. What emerged from these meetings was a sense of a vibrant field contributing to tools being increasingly applied in research across the social sciences. Given these developments, we thought it was timely to bring together a representative selection of articles that highlight some of the key issues and developments in this rapidly evolving field.
The first article, by Anselin, Syabri, and Kho provides an overview of GeoDa, a free program supporting interactive exploratory spatial data analysis. GeoDa is targeted at the non-GIS (geographical information systems) user and is designed to require as little as possible in terms of other software packages. Its goal is to provide analysts with a user-friendly, but powerful, integrated package that supports all phases of an empirical spatial data analysis, from mapping and geovisualization to spatial autocorrelation analysis, multivariate exploratory data analysis, and finally confirmatory spatial regression analysis. Although the implementation described in this article is a closed-source package, an ongoing research effort is porting the code to an open-source toolkit and making the package available to multiple platforms (Windows, MacOS X, and Linux/Unix).
In the second article, Roger Bivand provides an overview of recent work implementing functions for spatial statistical analysis in the open-source R language environment. Bivand draws on lessons learned from the R project about how to combine both substantive expertise and infrastructure support to achieve critical mass and mutually beneficial knowledge and tool exchange. The article describes how the R project functions in general, and outlines the main packages in the project designed for the analysis of spatial data. The spatial analytical capabilities of the spdep package are then illustrated using data on sudden infant death syndrome (SIDS) in North Carolina. Future opportunities for advancing spatial data analysis in R are also outlined.
Ned Levine's article, about the CrimeStat spatial statistics program, presents a description of the main components of the program and outlines its potential use both in crime-related research as well as in fields such as geography, epidemiology, forestry science, botany, and geology. CrimeStat is a stand-alone Windows program for the analysis of the spatial pattern of crimes and is designed to interface with most desktop GIS programs. The interface is carried out through using graphical objects produced by CrimeStat, which are linked to packages such as ArcView, ArcGis, and Maplnfo. Levine describes a number of examples involving spatial autocorrelation, hot-spot detection, spatial interpolation, and journey-to-crime analysis, which illustrate the statistical and visualization capabilities of the package.
While the first three articles in the issue primarily deal with area and point data, the article by Okabe, Okunuki, and Shiode outlines a GIS-based toolbox for network data, SANET (Spatial Analysis on a NETwork). …