Computational Simplifications Needed for Efficient Implementation of Spatial Statistical Techniques in a GIS

Abstract This paper contributes to the ongoing debate about which spatial analysis functions should be coupled with a GIS by identifying research problems that need to be solved before a richer toolbox of spatial statistical techniques can be implemented in a GIS. Three general problem areas are addressed. The first replaces a sequential ordinary least squares linear regression implementation with a single regression analysis. The second establishes the effective sample size for a single variable in a georeferenced data set, a result useful when calculating confidence intervals for means. The third establishes the effective sample size for pairs of variables in a georeferenced data set, a result useful when calculating the significance of correlation coefficients. These three general problems allow four more specific research problems to be identified that are in need of definitive solutions before a richer toolbox of spatial statistical techniques can be relatively easily implemented in a GIS. Their comp...