Some Extensions of Buser and Baroni‐Urbani's Clustering Method
L Dragomirescu · Biometrical Journal · 1987
Abstract We consider that the algorithm of numerical taxonomy focused upon the concept of “homogeneity”, proposed by Buser and Baroni‐Urbani (1982) for binary data, satisfies best from all the other algorithms known by us, the conditions of classification in biology formulated by Beckner in 1959. Therefore the paper proposes and studies homogeneities for binary and for ordered multi‐states data, which model explicitly Beckner's conditions. From all these, the homogeneity denoted by h* has the remarkable quality of generalizing the famous Jaccard's similarity coefficient. Moreover, using h* in this algorithm “naturally” improved by us, we can achieve the performance of a correct classification of the remarkable example proposed by Watanabe (1969).