Global Hot Spot Analysis (Getis-Ord G Statistic)

Both measure global spatial clustering, but Moran's I combines clustering of high values and clustering of low values into one index. The Getis-Ord G statistic separates these: a positive G signals that high values cluster together, while a negative G signals that low values cluster. This makes G more informative when the direction of clustering matters.

A common rule is to use the minimum distance at which every feature in the dataset has at least one neighbor, ensuring no isolated observations. Sensitivity analysis — running G at multiple distances — helps reveal whether the clustering finding is robust or an artifact of a particular threshold.

Use Global G first to confirm that spatial concentration exists across the study area. If G is significant, follow up with Local Gi* (or the local Getis-Ord Gi* tool) to pinpoint where the hot spots and cold spots are located. Running only the local version without the global test risks over-interpreting local fluctuations.

Yes, provided centroids (or population-weighted centroids) represent the polygon locations and attribute values are strictly positive. For polygon data, contiguity-based weights are also sometimes used, though the original G statistic was formulated with distance-band weights.

A non-significant G means there is no evidence of global concentration of either high or low values across the full study region. However, it does not rule out local clusters that cancel each other at the global scale. Always complement a non-significant G with a visual exploratory map before drawing conclusions.