Xiaoming Li1, Bonita Stanton, Shuli Yu. 1. Prevention Research Center, Carmen and Ann Adams Department of Pediatrics, Wayne State University School of Medicine, 4201 St. Antoine St., UHC 6-D, Detroit, MI 48201, USA. xiaoming_li@wayne.edu
Abstract
PURPOSE: The present study was designed to examine the underlying factorial structure of various problem behaviors among diverse populations using data from multiple cohorts of community-based samples of low-income urban African-American and rural Caucasian children and adolescents over multiple years. In this study, we tested the model of a single underlying problem behavior factor against competing models with multiple (and second-order) problem behavior factors. METHOD: The current study employed five data sets, four of which were collected in four consecutive community-based risk assessment or risk-reduction studies among urban low-income African-American children and adolescents in an eastern city in the United States from 1992-1996. The fifth was collected among rural primarily Caucasian adolescents living in Appalachia in 2000. Exploratory factor analysis with oblique rotation was performed to generate the factors underlying various adolescent problem behaviors. Confirmatory factor analysis was conducted based on the initial factors generated through the exploratory factor analysis to compare three competing models: single-factor model, multiple-factor model and one-factor second-order model. RESULTS: The data in the current study support the multiple-factor structure of adolescent problem behaviors. At the same time, the data also support the notion of a one-factor-second-order structure underlying various adolescent problem behaviors. The findings were robust across five data sets despite variations in: 1 samples (different cohorts with different demographics and different problem behavior profiles), 2) the types of problem behaviors examined, 3) the methods of data collection (e.g., computer assisted and paper and pencil), and 4) the number of problem behaviors and first-order factors involved. In addition, the results were also robust across gender and age groups. CONCLUSIONS: Compared with the single-factor model, the alternative models (i.e., multifactor model and one-factor second-order model) better explained the relationships among various measures of adolescent problem behaviors. The findings in the current study will help us for a better understanding of adolescent risk behaviors and contribute to more effective assessment and prevention intervention efforts.
PURPOSE: The present study was designed to examine the underlying factorial structure of various problem behaviors among diverse populations using data from multiple cohorts of community-based samples of low-income urban African-American and rural Caucasian children and adolescents over multiple years. In this study, we tested the model of a single underlying problem behavior factor against competing models with multiple (and second-order) problem behavior factors. METHOD: The current study employed five data sets, four of which were collected in four consecutive community-based risk assessment or risk-reduction studies among urban low-income African-American children and adolescents in an eastern city in the United States from 1992-1996. The fifth was collected among rural primarily Caucasian adolescents living in Appalachia in 2000. Exploratory factor analysis with oblique rotation was performed to generate the factors underlying various adolescent problem behaviors. Confirmatory factor analysis was conducted based on the initial factors generated through the exploratory factor analysis to compare three competing models: single-factor model, multiple-factor model and one-factor second-order model. RESULTS: The data in the current study support the multiple-factor structure of adolescent problem behaviors. At the same time, the data also support the notion of a one-factor-second-order structure underlying various adolescent problem behaviors. The findings were robust across five data sets despite variations in: 1 samples (different cohorts with different demographics and different problem behavior profiles), 2) the types of problem behaviors examined, 3) the methods of data collection (e.g., computer assisted and paper and pencil), and 4) the number of problem behaviors and first-order factors involved. In addition, the results were also robust across gender and age groups. CONCLUSIONS: Compared with the single-factor model, the alternative models (i.e., multifactor model and one-factor second-order model) better explained the relationships among various measures of adolescent problem behaviors. The findings in the current study will help us for a better understanding of adolescent risk behaviors and contribute to more effective assessment and prevention intervention efforts.
Authors: Lisa H Jaycox; Daniel McCaffrey; Beth Eiseman; Jessica Aronoff; Gene A Shelley; Rebecca L Collins; Grant N Marshall Journal: J Adolesc Health Date: 2006-07-10 Impact factor: 5.012