Abstract
Three methods for the analysis of spatial pattern, the Goodall (1974), the Greig-Smith (1952, 1964) and the Zahl (1974) S method are compared by Monte Carlo sampling and, for the latter 2 methods, on real data. For all 3 methods the data consists of counts or other measure of density in an array of plots. The analysis in each case tests the null hypothesis of independent, identically distributed plot variables against the multiple alternative of clusters of various sizes. Since analytical methods for determining critical values for the Greig-Smith and Goodall do not exist, these are determined empirically for both these methods for the Monte Carlo comparison. These values are quite stable over a wide range of distributions for Greig-Smith''s method, leading to a conjectured formula. The critical values are highly unstable for Goodall''s method, which is therefore not used for the real data. The estimated significance level of Zahl''s method is close to the nominal level over this same range of distributions. Greig-Smith''s method gives the best overall results, Goodall''s method the poorest results in the Monte Carlo sampling. The Greig-Smith and Zahl analyses of the real data result in similar conclusions about the extent and degree of clustering.

This publication has 1 reference indexed in Scilit: