MerQuR+Statistics Glossary

Statistics Glossary → Complete spatial randomness (CSR)

Complete spatial randomness (CSR)concept

The null hypothesis, defined by the homogeneous Poisson process, that points are scattered over the study area independently of one another and with the same intensity everywhere.

Under CSR, the number of points in any subregion follows a Poisson distribution whose mean is proportional to the subregion's area, and counts in disjoint regions are independent. In point pattern analysis the observed pattern is compared with CSR: if points are closer together than expected the pattern is clustered (e.g. plants that drop seeds nearby), and if they are more evenly spaced than expected it is regular or inhibited (e.g. the nests of territorial animals). Ripley's K and L, the nearest-neighbour distance distribution G, the empty-space function F and the J function are compared with their theoretical values under CSR; for example, K(r) = πr² under CSR.

Whether departures from the theoretical values are significant is usually assessed with Monte Carlo envelopes built from many patterns simulated under CSR, or with the Clark-Evans and quadrat count chi-square tests. Rejecting CSR does not explain why clustering occurs: it may come from interaction between points (a second-order effect) or simply from environmental conditions (soil, moisture) varying across the area, i.e. inhomogeneous intensity (a first-order effect). Distinguishing the two needs methods such as the inhomogeneous K function. The result also depends on how the observation window is chosen.

An everyday example

A park authority reviews where benches are placed. If the 50 benches were scattered completely at random, each 100 m² square would be expected to hold 0.5 benches on average. In the observed pattern most benches gather by the lake and the playground, with almost none on the open lawns; the quadrat count test rejects CSR with p < 0.001, and the pattern is clustered.

In MerQur

The basic null hypothesis of point pattern analysis; Ripley's K and L, g(r), the G, F and J functions, the Clark-Evans test, the quadrat count chi-square test and the Monte Carlo envelopes are all interpreted relative to CSR.

Analyses where this concept appears

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

Related concepts

Ripley's K/L and the G, F, J functionsMonte Carlo (simulation) envelopeQuadrat count test