Sampling distribution of consensus indices when all bifurcating trees are equally likely

Abstract

Consensus indices (CI) based on a consensus tree (CT) have been increasingly used in recent studies to measure the congruence between hierarchic classifications. But their statistical properties have yet not been investigated. This study reports on the distribution of CIs examined by a Monte Carlo method. The model employed assumes that all possible rooted bifurcating trees are equally likely. Other models involving random generation of data, similarity matrices, or other types of trees are under investigation.