Mapping Influence Regions in Heirarchical Clustering

Three-dimensional response surface plots arc presented for several hierarchical clustering methods and simulated core group data structures. The response surface represents the relative influence of the corresponding coordinate location in the bivariate data space on the clustering of the core groups. lnfluence is defined as a change in the clustering of the core groups when an element situated at a specified coordinate location is removed from the data set. The nature of the influence, whether beneficial or detrimental to the clustering process, can be determined from the plots. The influence plots revealed substantial differences between clustering methods. The influence patterns also explain some of the results found in previous validation research on clustering methods. The results have significant implications on the choice of clustering methods to be used in empirical research.