Development of Externalizing Behaviors in Chicago Youth Exposed to Intimate Partner Violence, Illinois, 1994-2002

Using data from all three waves of the Project on Human Development in Chicago Neighborhoods (PHDCN), this secondary data analysis examined the long-term effects of intimate partner violence (IPV) exposure during childhood and adolescence on subsequent externalizing behaviors (i.e., delinquency, violence, and substance use related offending). The research questions for this study were as follows: Are there significant differences in the mean scores of different externalizing behaviors (measured as "offending" in the present study) in any of the three PHDCN waves between youth exposed to IPV and youth not exposed to IPV? Are there distinct developmental trajectories of externalizing behaviors among youth exposed to IPV when compared to those not exposed to IPV? How do different individual- and neighborhood-level variables act in predicting the developmental paths of externalizing behaviors among youth exposed to IPV? Propensity score matching (PSM) was employed to match individuals reporting IPV exposure with those not exposed to IPV on key variables. Longitudinal latent class analyses (LLCA) were utilized to estimate the longitudinal developmental trajectories of externalizing behaviors independently for IPV and non-IPV exposed males and females and compared to each other. Multinomial logistic regression models were estimated separately for males and females exposed to IPV during their childhoods to examine the effect of different hypothesized class membership predictors. This collection contains a master dataset primarily sourced from Emery's (2006) data augmentation along with key variables from all three waves from the PHDCN Longitudinal Cohort Study, cohorts 12 and 15 (DS1); datasets constructed solely for multinomial logistic regressions for youth exposed to IPV, separated by sex (DS2 and DS3); data for the final LLCA models separated by sex and exposure to IPV (DS4 to DS7); and probabilities and latent classes created using Mplus (DS8 to DS9) that can be merged to the multinomial regression data using the <b>SUBID</b> variable. Additionally, syntax for variable and model constructions, as well as Mplus output, have been included as a zip package. Please refer to the P.I. documentation for more information.