On the Information Content of Classifications

Abstract

- The best fit estimate of the retrospective information content of a classification is argued to be the proportion of the maximal sum of informative subgroups to which each included taxon is assigned. The best fit estimate of the prospective information content of a classification is argued to be the proportion of fully resolved dadograms prohibited by that classification. Those (wo proportions are multiplied to provide a general information index that is sensitive to both the degree of resolution and the contents of individual subgroups. That index and the best fit considerations supporting it have implications for congruence studies, consensus indices, and the choice between Adams and Nelson consensus trees.