Blockmodels - Developments and Prospects (From Blockmodel Techniques for Criminal Justice Research - Final Report, Appendix B, 1981, by Phipps Arabie - See NCJ-88432)

• NCJ Number: 88433
• Author(s): P Arabie; S A Boorman
• Date Published: 1981
• Length: 31 pages
• Grant Number(s): 78-NI-AX-0142; SOC 76-24512; SOC 76-24394
• Sponsoring Agency: National Science Foundation; National Institute of Justice (NIJ); US Dept of Justice NIJ Pub
• Publication Format: Article
• Language: English
• Country: United States of America

Annotation

This paper discusses considerations of various algorithms and their implementations for obtaining blockmodels, user-oriented considerations, and implications for substantive theory and acquisition of appropriate data bases.

Abstract

The most widely used algorithm for obtaining blockmodels and image matrices has been the CONCOR (CONvergence of iterated CORrelations) algorithm. This procedure accepts continuously varying data and gives as output a dendogram, characteristic of the discrete output from hierarchical clustering. Hubert and Baker (1978) have pointed to recently proposed combinatorial clustering algorithms as potential rivals to CONCOR for obtaining blockmodels. Practical considerations discussed include the form of the input data, diagonals, goodness-of-fit, and others. The paper notes that evaluation problems are expected to arise as blockmodeling advances. Equations, figures, 5 footnotes, and about 70 references are included.