Robust Getis-Ord G and G* statistic devised by Julian Bruns (2018)12. The modified function reduces spatial scale effects by replacing the global parameter with a focal (regional) parameter. The method is implemented to extend the spdep-package by Roger Bivand
📍 disy Informationssysteme GmbH. https://www.disy.net/de/
🔧 spdep-package, for neighborhood definition and weighting schemes
As for spdep::localG(), additionally:
- Input: weighted neighborhood list ("listw" "nb"), defining the focal3 parameter listw_f
- Output: localG, list containing the geographically weighted z-scores of Focal Getis-Ord G* - can be converted to a vectors ton then be attached to a spatial class (package: sp)
The function is written in the same manner as spdep::localG, implementing the original Getis-Ord G* and can be applied in thevery same way. Additionally, the parameter listw_f has to be defined (spdep::poly2nb and spdep::nb2listw).
The area defined by the focal parameter should be smaller then the analysed area of study (global). For focal areas that include the analysed area of study as a whole, the obtained results become equal to the original Getis-Ord G*.
Exemplary visualization (ggplot2) of the resulting z-scores for focalG():
1 https://publikationen.bibliothek.kit.edu/1000083353
2 https://publikationen.bibliothek.kit.edu/1000071483
3 Focal Getis-Ord G* replaces the global mean of the original Getis-Ord G* with a focal mean that changes for every iteration, representing the focal neighborhood to that the current spatial analysis unit is compared to (in order to obtain statistical z-scores).