Adolescent risk behaviors and developmental contextualism: A person-oriented approach

Abstract

Issues surrounding person-oriented approaches to the study of risk behavior served as the impetus for this research. Cluster analysis is one such person-oriented technique that allows for the identification and empirical classification of individuals based upon a specified set of factors. However, issues remain as to the most appropriate cluster analytic strategies to use when studying adolescent risk behaviors. The first goal of this study was to explore differences found when clustering adolescent risk behaviors at time 1 and retaining the same groupings over time, compared with clustering at each point in time and examining the correspondence of the clusters. The second goal was to explore ways in which cluster analyses shed light on gender similarities and differences in the course and correlates of risk behaviors. Three waves of the National Youth Survey were used resulting in a final sample of 1391 (717 males, 674 females) individuals used in the cluster analyses. Using Ward's minimum-variance method, the optimal cluster solution yielded five clusters based upon the pseudo-T and consideration of group sample sizes. Results indicate differing patterns when individuals are allowed to re-cluster at each time period compared with maintaining the same clusters across time. Gender differences emerged; suggesting differential patterns associated with adolescent risk behaviors and the benefits of clustering separate samples of women and men. Implications for past and future research are discussed.