Complete-link clustering as a complement to factor analysis: A comparison to factor analysis used alone

Abstract

The rationale and method of the complete-link hierarchical clustering technique are discussed, along with some recently developed procedures for evaluating an overall clustering solution and for defining an “optimal” partition in a hierarchy. The application of factor analysis prior to clustering is suggested as a means of eliminating variables which do not conform well to the predominant structure in the data; thus, cluster analysis is used as a complement to factor analysis, capitalizing upon the strengths of both techniques. The advantages of the suggested method are then demonstrated using a previously published task analysis study of the roles and functions of the rehabilitation counselor. In comparing the obtained solution to the original factor analytic solution, it is argued that the two were generally consistent and that differences observed favored the clustering solution in terms of greater meaningfulness and homogeneity of the groupings. In addition, the clustering solution offered further advantages in the substantive interpretation of the data.