Second, we examined time-invariant predictors of ERO trajectories. We simultaneously examined the association of 3 binary measures of trauma exposure before 10 years (nonassaultive, non-sexual assaultive, and sexual assaultive traumatic exposures) with ERO intercepts and slopes (linear change from baseline through follow-up 4). This is depicted in Figure 1. Initial results indicated no evidence of nonlinear (ie, quadratic) effects. Modeling was conducted in Mplus 7.4 (Muthén and Muthén, 1998–2015) using full maximum likelihood estimation with robust standard errors. Age, gender (0 = male, 1 = female), and self-reported race/ethnicity (0 = non-Hispanic white, 1 = non-Hispanic black/African American, 2 = other) were used as covariates in all analyses. In addition, we accounted for genetic relatedness among siblings. Subsequent models included participants’ alcohol and cannabis use (0 = never used, 1 = ever used) at each interview, parental history of AUD, and participants’ impulsivity as measured by baseline BIS and SSS scores. Third, we evaluated whether residualized change in ERO from baseline to most recent follow-up was related to AUD, CUD, and INT and EXT psychopathology at each participant’s most recent interview.
