Combining Cluster and Discriminant Analysis to Develop a Social Bond Topology of Runaway Youth

Abstract

Typology development using discriminant analysis is extremely valuable in learning more about heterogeneous client populations of interest to social workers. However, the discriminant analy sis procedure is confined to situations in which individuals can be categorized into subgroups by the social work researcher. Classifying individuals by demographic variables such as gender or services utilized is straightforward and presents few problems. However, when more subjective categories are used to identify subgroup membership, problems may arise. There are many populations, such as the sexually abused, alcohol and drug abusers, sex offenders, and so on, who may avoid answering direct questions or deliberately give incorrect answers on questions formulated to classify them. This places in question the results of the discriminant analysis. In this article, data on 258 runaway and homeless youth are used to demonstrate how cluster analysis can be employed to create a classification variable needed to perform discriminant analysis. First, the data are submitted to cluster analysis to identify the groups in the population. Then, utilizing the cluster groups as the classifying variable needed to perform a discriminant analysis, the discriminantfunctions are extracted to develop a typology. The limitations of cluster analysis and this approach are discussed.