Abstract
Many techniques of numerical classification, as a method of multivariate analysis, have recently been developed. If different techniques are used for analysis, different classifications will be obtained; if different variates or different samples are used for classification, different results will be obtained. These results are also different from the classification by systematics. Hence a similarity measure for comparing classifications is required. A measure using BRILLOUIN'S (1962) information theory is proposed in this report for this purpose. This measure is modified slightly from a diversity measure which has been used by community ecologists such as MARGALEF (1958) and PIELOU (1974). The properties of this method are shown by numerical illustrations, and are compared with other methods such as those developed by RAND (1971) and BORKO et al. (1968).
