Scale and correlation in multiscale geographically weighted regression (MGWR)

Wei Kang & Taylor M. Oshan

Journal of Geographical Systems2025https://doi.org/10.1007/s10109-025-00468-1article
AJG 1
Weight
0.54

What the paper says

Multiscale geographically weighted regression (MGWR) extends geographically weighted regression (GWR) by allowing process heterogeneity to be modeled at different spatial scales. While MGWR improves parameter estimates compared to GWR, the relationship between spatial scale and correlations within and among covariates—specifically spatial autocorrelation and collinearity—has not been systematically explored. This study investigates these relationships through controlled simulation experiments. Results indicate that spatial autocorrelation and collinearity affect specific model components rather than the entire model. Their impacts are cumulative but remain minimal unless they become very strong. MGWR effectively mitigates local multicollinearity issues by applying varying bandwidths across parameter surfaces. However, high levels of spatial autocorrelation and collinearity can lead to bandwidth underestimation for global processes, potentially producing false local effects. Additionally, strong collinearity may cause bandwidths to be overestimated for some processes, which helps mitigate collinearity but may obscure local effects. These findings suggest that while MGWR offers greater robustness against multicollinearity compared to GWR, bandwidth estimates should be interpreted with caution, as they can be influenced by strong spatial autocorrelation and collinearity. These results have important implications for empirical applications of MGWR.

9 citations

Open paper page →

Cite this paper

https://doi.org/https://doi.org/10.1007/s10109-025-00468-1

Or copy a formatted citation

@article{wei2025,
  title        = {{Scale and correlation in multiscale geographically weighted regression (MGWR)}},
  author       = {Wei Kang & Taylor M. Oshan},
  journal      = {Journal of Geographical Systems},
  year         = {2025},
  doi          = {https://doi.org/https://doi.org/10.1007/s10109-025-00468-1},
}

Paste directly into BibTeX, Zotero, or your reference manager.

Flag this paper

Scale and correlation in multiscale geographically weighted regression (MGWR)

Flags are reviewed by the Arbiter methodology team within 5 business days.


Evidence weight

0.54

Balanced mode · F 0.40 / M 0.15 / V 0.05 / R 0.40

F · citation impact0.52 × 0.4 = 0.21
M · momentum0.72 × 0.15 = 0.11
V · venue signal0.50 × 0.05 = 0.03
R · text relevance †0.50 × 0.4 = 0.20

† Text relevance is estimated at 0.50 on the detail page — for your query’s actual relevance score, open this paper from a search result.