Parent, Peer, and Neighborhood Risk Factors Accounting for Spatial Clustering of Adolescent Aggressive Behavior Open Access
Downloadable ContentDownload PDF
The existing aggression literature highlights the importance of risk factors from multiple domains, including the parent, peer, and more recently, the neighborhood domains. However, this literature generally has not extended the study of neighborhood effects on adolescent aggression to include geographic location. Of the existing research examining neighborhood effects on youth aggression, neighborhood is typically measured using aggregate data of bounded areas, although aggressive behavior may vary according to geographic locations within pre-defined boundaries. Reliance only on aggregate data may lead to errors such as the ecological fallacy (i.e., the assumption that individuals have the same average characteristics of the larger group) or the overestimation or underestimation of model parameters (Diez-Roux, 2007). Thus, using analyses that link geographic information to individual data, or spatial analyses, may be particularly valuable given these problems. The present study used both spatial and non-spatial analyses to examine whether observations of high adolescent aggression and its risk factors geographically cluster, or occur in geographic proximity to other observations, as well as whether geographic clusters of adolescent aggression overlap or coincide with clusters of risk factors. Nearest Neighbor Analyses and multi-level analyses were conducted with a community, epidemiologically-defined sample of adolescents in 8th grade residing in Baltimore City. Results of spatial analyses revealed that aggression and its risk factors of low parental monitoring, high deviant peer affiliation, high community violence exposure, and high perceived neighborhood violence geographically clustered, and that these clusters overlapped to some extent. However, results of non-spatial analyses did not indicate clustering at the census tract level. Clinical implications of study results and future directions for using both types of analyses are discussed.