Parent-child relationship quality and parental autonomy granting were regressed on Wave 1 covariates and alcohol abuse was regressed on all Wave 1 and 4 covariates (not shown). Models were estimated with maximum likelihood estimation with robust standard errors to accurately test indirect effects and to account for missing data. For this reason, it was not possible to provide traditional fit indices such as chi-square test of model fit, TLI, CLI, RMSEA, and SRMR, or statistically test differences between nested and alternative models. The Mplus bootstrap option also does not work with this estimator nor the integration statement that are required to estimate indirect effects with weighted survey data. It was also therefore unable to provide bootstrapped confidence intervals for indirect effects.