Local Hot Spot Analysis (Getis-Ord Gi*)
Also known as: local Getis-Ord Gi*, Gi* statistic, spatial hot spot detection, local spatial clustering
Local Hot Spot Analysis uses the Getis-Ord Gi* statistic to identify specific geographic locations where high or low values cluster together more than expected by chance. Unlike global measures that return a single summary for the whole study area, this local statistic produces a z-score for each feature, pinpointing exactly where statistically significant hot spots and cold spots occur.
Read the full method
Sign in with a free account to read this section.
Method map
The neighbourhood of related methods — select a node to explore.
When to use it
Use Local Hot Spot Analysis when you need to identify specific geographic locations — not just detect whether clustering exists globally — where high or low values concentrate significantly. It is appropriate for georeferenced continuous data such as crime counts, disease incidence rates, property values, or environmental measurements, and when the research question demands actionable, place-specific findings. Avoid it when the dataset has very few observations (fewer than ~30), when attribute values are binary rather than continuous, when the spatial weights specification is theoretically unjustifiable, or when a single global answer suffices and location-specific detail is not needed.
Strengths & limitations
- Produces a significance map: every location receives its own z-score, enabling precise identification of where clusters occur.
- Distinguishes hot spots (high-value clusters) from cold spots (low-value clusters) in a single analysis.
- Robust to irregularly shaped study areas and works with both point and polygon data.
- Widely implemented in GIS software (ArcGIS, GeoDa, R spdep) with well-documented inference procedures.
- FDR correction and Monte Carlo permutation options make inference reliable under non-normality.
- Results are sensitive to the spatial weights specification; different distance bands or neighbor counts can change which locations appear significant.
- Requires a sufficiently large sample — small datasets yield unstable z-scores and inflated false-positive rates.
- The Gi* statistic cannot distinguish between a true spatial process and a global trend (non-stationarity) confounding local patterns.
- Does not identify spatial outliers (a high value surrounded by low values); Local Moran's I is needed for that distinction.
Frequently asked
What is the difference between Local Hot Spot Analysis (Gi*) and Local Moran's I?
Both are local spatial statistics, but Gi* identifies clusters of similarly high or low values (hot and cold spots), while Local Moran's I also detects spatial outliers — locations where a high value is surrounded by low values or vice versa. For pure hot spot detection, Gi* is preferred; for a fuller picture of spatial association types, use both.
How do I choose the spatial weights or distance threshold?
The threshold should reflect a theoretically meaningful scale of interaction — for example, a realistic travel distance for crime, or the maximum dispersal range for a species. Use an incremental spatial autocorrelation analysis to find the distance at which clustering peaks, or examine a semivariogram for continuous data.
Why should I apply a false discovery rate (FDR) correction?
Running a significance test for every feature in a large dataset creates many opportunities for false positives. FDR correction limits the expected proportion of false discoveries, making your hot spot map more reliable. ArcGIS applies FDR by default; GeoDa and R require it to be set explicitly.
Can Local Hot Spot Analysis handle population-weighted count data?
Yes. For count data such as disease cases, you should first compute a rate (cases per population) or use spatial filtering to adjust for population density before running Gi*. Failing to do so produces hot spots that simply reflect dense areas rather than genuinely elevated risk.
Is a permutation test better than the analytical z-score?
The analytical z-score assumes approximate normality, which may not hold with skewed data or irregular polygon shapes. A conditional permutation test (available in GeoDa and R spdep) randomizes attribute values across locations to simulate the null distribution, giving more reliable p-values for non-normal data.
Sources
- Ord, J. K., & Getis, A. (1995). Local spatial autocorrelation statistics: Distributional issues and an application. Geographical Analysis, 27(4), 286-306. DOI: 10.1111/j.1538-4632.1995.tb00912.x ↗
- Getis, A., & Ord, J. K. (1992). The analysis of spatial association by use of distance statistics. Geographical Analysis, 24(3), 189-206. DOI: 10.1111/j.1538-4632.1992.tb00261.x ↗
How to cite this page
ScholarGate. (2026, June 3). Local Hot Spot Analysis (Getis-Ord Gi*). ScholarGate. https://scholargate.app/en/spatial-analysis/local-hot-spot-analysis
Which method?
Set this method beside its closest kin and read them side by side — the library lays the books on the table; the choice is yours.
- Hot Spot AnalysisSpatial analysis↔ compare
- Local Getis-Ord Gi*Spatial analysis↔ compare
- Local Indicators of Spatial AssociationSpatial analysis↔ compare
- Local Moran's ISpatial analysis↔ compare
- Local Spatial AutocorrelationSpatial analysis↔ compare