Spatial econometrics in an age of CyberGIScience

Abstract

In this article, we focus on the evolution of the technology that lies at the basis of implementing spatial econometric methods into software tools. We review the changing methodological emphases and their implications for data structures and computational infrastructure required for estimation and inference. We review the evolution of software solutions, starting with SpaceStat and GeoDa and moving on to the PySAL open source library of spatial analytical functions (Rey and Anselin 2010, PySAL: a Python library of spatial analytical methods. In: M.M. Fischer and A. Getis, eds. Handbook of applied spatial analysis. Berlin: Springer, 175–193.). We compare these approaches with other software solutions, such as the R spatial analytical routines and recently released Stata functionality for spatial econometrics. We follow the review with a discussion of requirements and challenges encountered when moving these software tools into a CyberGIScience framework. We focus on the efficient data structures, the need for metadata and provenance tracking, as well as high-performance computing requirements. We close with the outline of a vision for a ‘spatial econometrics workbench’ as a core component of cyberinfrastructure for GIScience.

Keywords:

• spatial econometrics
• cyberinfrastructure
• cyberGIS
• GIScience

Acknowledgements

This research was funded in part by NSF Award OCI-1047916, SI2-SSI: CyberGIS Software Integration for Sustained Geospatial Innovation. We thank the anonymous referees and the editors for their constructive comments.

Notes

1 For overviews see Belsley and Kontoghiorghes (2009); Jacho-Chávez and Trivedi (2009); Hajivassilou (2008); Renfro (2004a).

2 As an illustration, consider an analysis of the contiguous US counties, with an n of roughly 3000. The associated weights matrix would contain 9,000,000 elements.

3 The original SpaceStat has no relationship to the current SpaceStat 3.0 distributed by Biomedware (http://www.biomedware.com).

4 As of this writing, the PySAL library does not yet contain maximum likelihood estimation.