A “Natural” Agglomerative Clustering Method for Biology

Summary The authors consider that cluster analysis does not objectivize but represents the biologist’s subjectivity as to (1) characters considered to be significant and to (2) the way of classification. The latter, however, authors’ opinion, must be specific to the field of application. To this effect some methods are suggested for biology. The methods originate in improvements or transformation of Buser and Baroni-Urbani’s method, as well as Watanabe’s method, and have the property of processing overall information with no loss or distortion. An agglomerative method which yields a necessarily unique result is suggested, being considered by the authors as a homologue of Watanabe’s divisive method. The methods proposed are studied using examples logically constructed. These examples can provide from biology, especially from ecology. Key words : Cluster analysis; Statistics of binary data; Subadditive measure; Ecology. 1. Introduction This paper is intended to present some hierarchical clustering methods which hold, in our opinion, a privileged position at least with respect to applications in biology. The superiority of these methods arises from their quality of being