MerQuR+Statistics Glossary

Statistics Glossary → Clark-Evans nearest-neighbour index

Clark-Evans nearest-neighbour indexconcept

A ratio that compares the average distance from points to their nearest neighbours with the distance expected under a completely random pattern, indicating whether the pattern is clustered, random or regular.

For each point the distance to its nearest neighbour is measured and averaged (r̄_o). Under a homogeneous Poisson process with intensity λ (complete spatial randomness), the expected mean distance is 1 / (2√λ). The index is the ratio R = r̄_o / r̄_e: R ≈ 1 indicates a random pattern, R < 1 a clustered one and R > 1 a regular (mutually avoiding) one; a perfectly regular hexagonal arrangement reaches R of about 2.15. The significance of the difference is tested with a z test.

The index is sensitive to the boundary of the study area: the true nearest neighbour of a point near the edge may lie outside the area, so distances are overstated, which is why an edge correction is applied. Because intensity is computed from the size of the area, the choice of the area's boundary also affects the result. Since it looks only at the nearest neighbour, it summarizes the pattern at a single scale; for patterns that behave differently at different distances, Ripley's K or L functions are more informative.

An everyday example

The positions of 100 trees are mapped in a 1-hectare plot of forest. If they were randomly scattered, the average distance to the nearest neighbour would be 5 metres; the measured average is 3 metres. With R = 0.6, the trees grow in a clustered way, perhaps in clumps because seeds fall around the parent tree.

In MerQur

Offered as the Clark-Evans test in Point Pattern Analysis, alongside the Ripley K/L, G, F and J functions.

Analyses where this concept appears

Point Pattern Analysis (Ripley K/L, G/F/J, CSR)

Related concepts

Complete spatial randomness (CSR)Edge (boundary) correction