Multivariate Methods in Plant Ecology: V. Similarity Analyses and Information-Analysis

Publisher Website
Google Scholar
Abstract
Agglomerative-polythetic methods (commonly known as "similarity methods") of hierarchically classifying elements into sets can take a large number of different forms, according to the type of fusion strategy ("sorting method") employed and the coefficient used to measure similarity. Ten selected versions, using 2 different sorting methods combined in turn with 5 different coefficients, are tested empirically for their relative efficiency, using both theoretical and ecological criteria. The results from the com- parative analyses of 2 test-communities show that, whereas ''centroid sorting in general gives better results than ''nearest neighbor sorting, there is also an interaction between sorting strategy and coefficient. The method combining centroid sorting with an information-statistic coefficient is shown to be greatly superior to the others in producing clear-cut and ecologically acceptable hierarchies; and this method, called information analysis, is selected for further test.