Some Final Comments and Guidelines
Barry J. Everitt, Sabine Landau, Morven Leese, Daniel R. Stahl · Wiley series in probability and statistics · 2011
This chapter gives an overview of the steps in a typical analysis. It discusses those steps concerned with cluster validation and interpretation, and two applications that illustrate many of the issues involved. A test for the absence of cluster structure may not be necessary if the reason for clustering is practical. Comparing partitions or trees, either with each other or with data, is a common requirement in cluster validation. Internal cluster quality can be taken to refer to the extent to which clusters meet the requirements for good clusters, as defined by Cormack, namely isolation and cohesion. Robustness refers to the effects of errors in data or missing observations, and changes in the data or methods. Controlled Vocabulary Terms robustness