ABSTRACT
The final chapter offers some reflections on the limits of spatial data exploration under an abductive approach. The difference between discovering interesting patterns and their explanation in terms of spatial processes is highlighted.
Three broader issues are considered: the potential pitfalls of data science and their implications for scientific reasoning, the limitations intrinsic to spatial analysis and the need for reproducible research. In an exploratory approach, one has to be conscious of the many researcher degrees of freedom, such as the selection of spatial scale, outliers, imputation, spatial weights and the choice of various tuning parameters. The importance of sensitivity analysis is stressed.
Spatial analysis has its own potential pitfalls in the form of ecological fallacy, the modifiable areal unit problem, the change of support problem and the importance of spatial scale. These limitations do not invalidate the results of data exploration, but they may affect their interpretation.
Finally, the need for reproducible research is stressed, and the limitations of GeoDa in that respect are assessed. The Appendix introduces the RGeoDa and PyGeoda packages that allow GeoDa functionality to be included in a scripting framework.
