Concerning the NJ algorithm and its unweighted version, UNJ
Olivier Gascuel · DIMACS series in discrete mathematics and theoretical computer science · 1997
In this paper we will present UNJ, an unweighted version of the NJ algorithm (Saitou and Nei 1987; Studier and Keppler 1988). We will demonstrate that UNJ is well suited when the data are of the ( ) ( ) d e ij ij ij d = + type, where ( ) d ij is a tree distance, and when the e ij are independent and identically distributed noise variables. Simulations confirm this theory. On a more general level, we will study the three main components of the agglomerative approach, applied to the reconstruction of tree distances. (i) We will