COBLOC: A hierarchical method for blocking network data

A method is described for finding blockmodels of a given network using an agglomerative hierarchical clustering technique with optional backtrack capabilities. The clustering criterion is an index of the goodness of fit of the blockmodel that would be produced by a given clustering. This index is computed for each clustering considered. This method is compared on a priori grounds with other methods and comparative results are reported for its application to the Western Electric Bank Wiring Room data. Extensions combining this method with other clustering methods are discussed.