Are “Dynamic” Predictors of Youth Violence Actually Dynamic?

Date

2015-08

Journal Title

Journal ISSN

Volume Title

Publisher

Abstract

Youth violence is a serious social problem with a 12-month prevalence rate of about 35 percent (Herrenkohl, Lee, & Hawkins, 2012). While research has identified dynamic predictors of violence, there is little evidence of their malleability and impact on youth violence since experimental studies are scarce and few correlational studies have examined within-individual differences. Also, few studies have applied item response modeling (IRM), which allows differential weighting of violent acts. The current study is the first to use multilevel modeling (MLM) to examine predictors of within-individual change in violence among males in the National Longitudinal Study of Adolescent Health (Add Health) data set. Due to sex differences in the rates of violent offending, the sample is restricted to males. It is only the second study to use IRM to scale the violence outcome measure. The sample includes males (N=2288) from the Add Health public dataset, which captures violent offending from Wave 1 (age 11-21) to Wave 2 (age 12-22). Samejima’s (1997) graded response model translated self-reported violence onto a continuous scale. MLM examined dynamic predictors of within-individual change in violence, static predictors of between-individual differences, and the interaction between age and peer delinquency. The IRM results showed that items varied in difficulty, poor factor loading for one item, and local dependence for two other items. The results of MLM indicated that, on average, individuals became less violent with age; Peer delinquency, a daily family meal, and alcohol use significantly predicted within-individual change in violence; and demographic variables, GPA, school attachment, history of grade retention, depressive symptoms, peer delinquency, and a daily family meal significantly predicted the level of violence between individuals.

Description

Keywords

Youth Violence, National study, Longitudinal study, Adolescent health, Multilevel modeling, Item response modeling, Violent offending

Citation