Due to the dichotomous nature of the outcome variable (adolescent lifetime substance use), we employed weighted least squares estimate with a mean and variance (WLSMV) adjusted χ2 statistic for all SEM path analyses. WLSMV supplies standard errors, unbiased estimates, and model fit tests for dichotomous outcome variables and for non-normal data in general (Muthén, 1984; Muthén & Muthén, 2006). We evaluated model fit using the weighted root-mean square residual (WRMR) statistic; values falling below .90 indicated a good model fit (Muthén & Muthén, 2006; Yu & Muthén, 2001). The WRMR is the only model fit statistic available when examining dichotomous outcome variables (Muthén & Muthén, 2006).
