Cluster Analysis in Sociometric Research: A Pattern-Oriented Approach to Identifying Temporally Stable Peer Status Groups of Girls

• NCJ Number: 217421
• Journal: Journal of Early Adolescence Volume: 27 Issue: 1 Dated: February 2007 Pages: 90-114
• Author(s): Peter Zettergren
• Date Published: February 2007
• Length: 25 pages
• Publication Format: Article
• Publication Type: Report (Study/Research)
• Language: English
• Country: United States of America

Annotation

This study explored the usefulness of a clustering procedure for sociometric classification purposes by identifying the peer status clusters of a sample of 445 girls in Sweden and analyzing their stability over time.

Abstract

The analysis identified temporally stable clusters of rejected, popular, and average girls, which were similar with the corresponding peer status groups identified by the frequently used Coie, Dodge, and Coppotelli (CDC) classification method. All of the girls in the rejected cluster were also selected as rejected by the CDC model and all CDC-identified stably rejected girls were also members of the current rejected cluster. However, the rejected and popular clusters were at least twice as large as the corresponding CDC peer status groups, suggesting that cluster analysis could overcome some of the problems of traditional sociometric methods in the identification of large homogeneous groups. As such, the findings support the use of the clustering technique as a classification method in longitudinal sociometric research. Data were drawn from a large-scale longitudinal Swedish research program. Participants were a cohort of all children attending Grade 3 in 1965 who were followed from age 10 onward. The current analysis focused on a sample of 445 girls with available peer data from age 10 and age 13. Participants were given a same-sex class roster and told to rank order girls in terms of who they wanted to remain with the class if the class was moved to a smaller classroom where there would not be room for everyone. Data were also gathered about their family socioeconomic status, academic achievement, and teacher ratings of their behavior. Cluster analysis was performed using a modified LICUR (Linking of Clusters after Removal of a Residue) procedure from the statistical package SLEIPNER. Results obtained from the cluster analysis were then compared to the results obtained from the CDC method. Future research with cluster analysis should examine different age periods from early childhood on and should use a more modern dataset than the one used for the current analysis. Tables, references