Building Better Blockmodels: A Non-Hierarchical Extension of CONCOR With Applications to Regression Analysis

Abstract

Cluster analysis has been used by many as a tool for identifying relatively small numbers of categories which represent the salient contrasts within bodies of data. One method which has received widespread use among sociologists in general, and sociometrists in particular, is CONCOR. This paper presents a non-hierarchical extension of the CONCOR algorithm and demonstrates the advantages of this extension with three exemplary bodies of data. One of the bodies of data is then. used as a vehicle for expanding on the information which is customarily derived from cluster, or blockmodel, analyses. Specifically, a means by which the results of blockmodeling may be translated into a regression model is described and demonstrated using these network data.