MerQuR+Statistics Glossary

Statistics Glossary → Local indicators of spatial association (LISA) and Gi*

Local indicators of spatial association (LISA) and Gi*concept

Local statistics that break global spatial autocorrelation down to individual locations, showing whether each unit forms a significant cluster of high or low values with its surroundings.

Global Moran's I gives a single number for the whole study area. Anselin's local Moran computes, for each unit i, Iᵢ ∝ zᵢ Σⱼ wᵢⱼ zⱼ, where zᵢ is the unit's standardized value and the sum covers the weighted values of its neighbours. A positive Iᵢ means the unit resembles its neighbours (a high-high or low-low cluster), a negative one that it differs from them (a high-low or low-high outlier). The local values sum to a quantity proportional to the global I. Getis-Ord Gi* instead compares the sum of values in a neighbourhood that includes the focal unit with the overall mean and gives a z score: a large positive z marks a hot spot, a large negative z a cold spot.

Inference is usually based on conditional permutation. Because a separate test is run for each unit and neighbouring tests are not independent, many false positives are expected; corrections such as the false discovery rate (FDR) are recommended. The results are sensitive to how the spatial weights matrix is defined (contiguity, distance band).

An everyday example

Burglary rates are examined in 120 neighbourhoods of a city. Global Moran's I is positive and significant but does not say where the clustering is. LISA shows that 8 neighbourhoods east of the centre form a high-high cluster and that one high-rate neighbourhood is surrounded by low-rate ones (a high-low outlier); Gi* flags the same eastern area as a hot spot at 99% confidence.

In MerQur

'Local I_i (LISA) definition' and 'LISA p definition (z_sim normal approximation, Folded permutation p)' in Moran's I; the 'Gi* / Gi' and 'Gi_Bin' checkboxes in Getis-Ord Gi* Hotspot (confidence classes ±3, ±2, ±1); and in the KDE Density and Hexbin maps by choosing Moran's I or Getis-Ord Gi* from the 'Method' list.

Analyses where this concept appears

KDE Density MapHexbin MapMoran's I (Spatial Autocorrelation)Getis-Ord Gi* Hotspot

Related concepts

Spatial autocorrelation (Moran's I, Geary's C)False discovery rate (FDR, Benjamini-Hochberg, Storey q)