Robust Geary's C
Also known as: robust Geary contiguity ratio, outlier-resistant Geary's C, robust spatial contiguity statistic, robust Geary C
Robust Geary's C adapts the classical Geary contiguity ratio — a measure of spatial autocorrelation based on pairwise squared differences between neighbouring locations — to resist distortion by spatial outliers and influential observations. It retains the local sensitivity of Geary's C while producing more reliable inferences when the spatial data contain extreme values or non-normal distributions.
Key highlights
- Resistant to distortion by spatial outliers and extreme values that inflate squared differences in the classical statistic.
- Retains Geary's C's sensitivity to local spatial dissimilarity, which can detect patterns that Moran's I misses.
- Permutation-based inference avoids assumptions of normality, making it valid for skewed or heavy-tailed spatial distributions.
- Produces more reliable global summaries of spatial structure in real-world datasets that routinely contain anomalous observations.
- Conceptually straightforward extension of a widely understood classical statistic, easing communication of results.
Intuition
This section is available to Pro members. Upgrade to Pro
How it works
This section is available to Pro members. Upgrade to Pro
When to use it
Use Robust Geary's C when you have areal or point-referenced continuous data and need to test for spatial autocorrelation but suspect the dataset contains spatial outliers, extreme values, or non-normal distributions that could bias the classical Geary's C. It is especially appropriate in applied settings such as disease mapping, crime analysis, or environmental monitoring where a few anomalous locations are common. Prefer the classical Geary's C when data are well-behaved and no outliers are present, as the robust variant adds complexity without benefit. Do not use either version when spatial stationarity is implausible across the study region — consider local statistics (Local Geary's C or LISA) instead.
Strengths & limitations
- Resistant to distortion by spatial outliers and extreme values that inflate squared differences in the classical statistic.
- Retains Geary's C's sensitivity to local spatial dissimilarity, which can detect patterns that Moran's I misses.
- Permutation-based inference avoids assumptions of normality, making it valid for skewed or heavy-tailed spatial distributions.
- Produces more reliable global summaries of spatial structure in real-world datasets that routinely contain anomalous observations.
- Conceptually straightforward extension of a widely understood classical statistic, easing communication of results.
- No single universally agreed robust formulation exists; different implementations (trimming, M-estimation, median scaling) can yield different results.
- Interpretation of the robust C value is less standardised than for the classical statistic, complicating cross-study comparisons.
- The robustness gain can mask genuinely extreme spatial processes that are substantively meaningful rather than mere data errors.
- Permutation inference is computationally more intensive than normal approximation tests, especially for large datasets.
Common pitfalls
This section is available to Pro members. Upgrade to Pro
Applications
This section is available to Pro members. Upgrade to Pro
Frequently asked
How does Robust Geary's C differ from classical Geary's C?
Classical Geary's C sums squared differences between neighbouring pairs, making it sensitive to extreme values. The robust version moderates the contribution of outlying pairs through trimming, median-based scaling, or influence-function methods, so extreme observations have less leverage over the overall statistic.
How does Robust Geary's C relate to Moran's I?
Both measure global spatial autocorrelation, but Geary's C focuses on pairwise local differences (values below 1 = positive autocorrelation), while Moran's I uses cross-products of deviations from the mean. Geary's C tends to be more sensitive to local dissimilarity, and its robust variant is preferable when local outliers are the concern.
What spatial weights matrix should I use?
Common choices include queen or rook contiguity for regular grids, k-nearest neighbours for irregular point patterns, and distance-band weights for continuous data. Run a sensitivity analysis using at least two different weights schemes to confirm your conclusions are not driven by the neighbour definition.
When should I prefer Local Geary's C over the global robust version?
Use the global robust Geary's C when you need a single summary statistic for the entire study region. Switch to Local Geary's C (or LISA) when you want to identify which specific locations drive spatial clustering or dissimilarity, or when spatial structure is not uniform across the region.
Is a permutation test always necessary for inference?
For the robust variant, permutation-based inference is strongly recommended because the robust transformation alters the null distribution in ways that the classical normal approximation does not account for. Permutation tests make no distributional assumptions and are valid for both robust and classical statistics.
Sources
- 1.Geary, R. C. (1954). The contiguity ratio and statistical mapping. The Incorporated Statistician, 5(3), 115–145.
- 2.Anselin, L. (1995). Local indicators of spatial association — LISA. Geographical Analysis, 27(2), 93–115.
You have read it. What now?
Cite this page
ScholarGate. (2026, June 3). Robust Geary's C. ScholarGate. https://scholargate.app/spatial-analysis/robust-gearys-c