correlation. Evidence for C arises from rDZ > ½ rMZ, while support for E is based on rMZ < 1. All our results are based on formal model-fitting procedures where raw data were fitted using full information maximum likelihood estimation in the statistical software package, Mx.33 Thresholds were adjusted for age at Wave 4 (binary, age ≥22; median split). Analyses were conducted in White and Black twins separately; where there were effects in both White and Black twins, parameters across racial/ethnic groups were compared using multi-group post hoc tests that examined whether constraining parameters across racial/ethnic groups resulted in a change in model fit.
