The bivariate model is shown with unstandardized path coefficients in Figure 1. The proportion of total variance attributable to genetic and environmental sources is reported with 95% confidence limits in Table 3. Equating A, C, and E across females and males did not produce a significant change in model fit for timing of first use (Δχ2 (3)=0.64; p=0.89) or for AD (Δχ2 (3)=0.27; p=0.97). Covariance in timing of first use and AD could also be equated for females and males without a deterioration in model fit (Δχ2 (2)=0.28; p=0.87). Dropping the C pathway for AD also did not result in a significant change in model fit (Δχ2 (2)=1.03; p=0.60), but C could not be dropped for timing of first use (Δχ2 (2)=6.98; p=0.03). Dropping the E pathway between the two phenotypes did not significantly impact model fit (Δχ2 (2)=3.88; p=0.14), indicating that although a portion of covariance was attributable to individual-specific factors common to the two phenotypes (16.5%), the contribution of common E was not statistically significant. By contrast, the A pathway representing the genetic covariance between timing of first alcohol use and AD could not be dropped without a deterioration in model fit (Δχ2 (2)=6.57;p=0.04).
