Regressão linear geograficamente ponderada em ambiente SIG

Abstract

This paper discusses theoretical issues and results of the implementation in a GIS environment of a spatial statistic confir-matory model — geographically weighted linear regression (GWR) — not available in a free software environment. The theoretical aspects of this local spatial regression model were vastly discussed due to the lack of related bibliography. The GWR model was coded in TransCAD’s programming language, named GISDK, making a comprehensive use of manipulation tools, data geo referenced treat-ment, and spatial analysis routines already existing in GIS platforms. At the end, we hope to have developed, although in a partial man-ner, an important tool which will contribute to a better understanding and fine tuning of geographical phenomena modeling so widely applied in transportation planning studies.