Using cluster analysis to assess the effects of stressful life events: Probing the impact of parental alcoholism on child stress and substance use

Cluster analysis was used to categorize stressful life events into groups based on the effect (or mediational) pattern linking each event to external criteria defined within an overarching conceptual model. To illustrate the utility of "effect size clustering," we examined the structure of stressful life events that may function to mediate, in part, the relation between parental alcoholism, parental psychopathology, and adolescent substance use. The analyses were conducted by forming clusters based on how each of 29 individual stressors related with 13 other variables considered relevant to the mediational process under investigation, rather than using the inter-item correlation matrix to determine event similarity. Using data on 326 families, four clusters were obtained that were labeled (a) family-related conflict, (b) general child relationship problems, (c) parent problems, and (d) major illness/ bereavement. These exploratory analyses revealed that the family-related conflict cluster of events were of key importance in accounting for the link between paternal alcoholism and child substance use. Advantages and limitations of using effect size clustering for examining life event data, as well as data used in other community-oriented and program evaluation applications, are discussed.