Troubles with trees
by
Roger Marshall
Dept Community Health University of Auckland

A critique is presented of the use of tree-based partitioning algorithms to formulate classification rules and identify subgroups from data. It is argued that the methods have a number of limitations, despite their popularity and apparent affinity to logical reasoning. The issue of redundancy in tree derived decision rules is discussed. Simple rules may be unlikely to be "discovered" by tree growing. Subgroups identified by trees are often hard to interpret or believe and net effects are not assessed. These problems arise fundamentally because trees are hierarchical. Newer refinements of tree technology seem unlikely to be useful, wedded as they are to hierarchical structures. Example from clinical and epidemiological data are presented.

Date received: May 6, 2002

Copyright © 2002 by the author(s). The author(s) of this document and the organizers of the conference have granted their consent to include this abstract in Atlas Conferences Inc. Document # cajj-04.