The present study uses data from individual twins to examine phenotypic associations between timing of substance involvement and both parental separation and parental alcoholism. Analyses were performed in STATA version 12 (StataCorp, 2011), with the Huber-White robust variance estimator used to compute standard errors and confidence intervals adjusted for non-independence (i.e., the correlated nature) of twin-family data. Comparisons of EA and AA families were conducted as part of preliminary analyses only. After confirming well-documented racial/ethnic differences, for example, in patterns of substance use during adolescence (Johnston et al., 2012) and family structure (Raley and Bumpass, 2003; Bramlett and Mosher, 2002), survival and propensity score analyses were conducted separately for EA and AA families. Rather than conducting pooled analyses modeling interactions with race/ethnicity, we were compelled to stratify because of the potential (and subsequently documented) problem of covariate inbalance, i.e., the existence of combinations of covariate and outcome values present in EA families but absent in AA families—under such conditions, a pooled analysis can lead to biased estimates (Rosenbaum and Rubin, 1983; Rubin, 2006).
