Direct Factor Analysis of Sociometric Data

Abstract

Electronic computation allows cliques or subgroups within a larger group to be separated by direct factor analysis of sociometric choices, followed by analytic rotation. Information on the positions of individuals within subgroups is also provided by this procedure, which is illustrated here with a group numbering 67. Existing computer routines permit extension to groups as large as 125; the principles may be extended to larger groups by additional programming. Most previous work on the identification of cliques or subgroups has centered about the linkage structure of the larger group; proposed methods have included graph theory (9, 10, 11) and raising the sociomatrix to higher powers (17). Computer programming of these methods for large groups becomes difficult because in principle they require storage capacity that increases rapidly with the number of links in the graph, or with the power of the matrix. It has also been proposed that rows and columns of the sociomatrix be permuted so as to bring choices nearer to the main diagonal (1), but computer programming of several methods of this type has shown that they do not always reveal subgroups satisfactorily (6). These difficulties can be circumvented by factorial examination of similarity of choices given or received by group members, a method used by Bock and Husain (2) and Gouldner (8), and discussed by Glanzer and