Getis-Ord Gi*: Finding Statistically Real Hot Spots
Any map has highs and lows, but which clusters are real and which are noise? The Getis-Ord Gi* statistic separates genuine spatial concentrations — hot spots and cold spots — from random chance, with a confidence level attached.
The physical basis
For each feature, Gi* compares the sum of its value and its neighbours' values against what you would expect if values were distributed randomly, returning a z-score. Large positive z-scores mark clusters of high values (hot spots); large negative z-scores mark clusters of lows (cold spots); the p-value tells you how confident to be.
The formula
Reading the values
> +1.96 significant hot spot · < −1.96 significant cold spot · in between = not significant (95% confidence).
What it runs on
| Operates on | A point or polygon layer with a numeric attribute — not tied to specific spectral bands. |
In practice
It is used in epidemiology (disease clusters), crime analysis, and retail-demand and hazard mapping — anywhere you need defensible clusters rather than eyeballed ones.
Where it struggles
Results hinge entirely on how you define "neighbours" (distance band or contiguity), so justify that choice; and with many features, correct for multiple testing (for example a false-discovery-rate adjustment) to avoid spurious hot spots.
Compute Getis-Ord Gi* on your own study area
Skip the code. Draw or upload a boundary and Spatial Research Suite runs Getis-Ord Gi* on live data — with cloud masking, exports and citations built in.
Run this analysis in GISforus →Frequently asked
What does the Gi* z-score mean?
It measures how strongly high (or low) values cluster around a feature: above +1.96 is a significant hot spot, below −1.96 a significant cold spot, at 95% confidence.
Why does the neighbourhood definition matter?
Gi* is computed over neighbours, so the distance band or contiguity rule you choose directly determines which clusters appear.