Order-neighbour analysis

Order-neighbour analysis analyses point patterns, and offers tests of the hypothesis of randomness of the pattern. Deviations from randomness can be identified as either clustered or dispersed. Mean distances between pairs of points are compared to those expected under the assumption of randomness, for which statistical tests are available, and also with observed values for non-random patterns. A major problem is the boundary effect: values are calculated for finite bounded areas, whereas, theoretical expectations are observed for infinite unbounded areas. Corrections can be made either by the use of 'buffered zones', by comparison with bounded random patterns, or else by the use of correction equations derived from certain restricted cases. A more detailed analysis of the patterns is given by the cumulative frequency distribution of inter-point distances, which can again be subject to tests of the hypothesis of randomness. The techniques are illustrated by a study of the location patterns of a number of activities within the cities of Sydney and Melbourne, Australia, and a critique offered of some published uses of the techniques.-A.Armstrong