their initial latent liability for alcohol abuse at grade 7 (σ̂2 = 7.79, SE = 1.32, p < .001) and rates of linear increase (σ̂2 = 0.35, SE = 0.06, p < .001). The intercept and linear slope were significantly correlated with each other, r = −0.64, SE = 0.05, p < .001. Therefore, adolescents who had lower initial liability for alcohol abuse at grade 7 tended to show higher rates of linear slope. In the conditional model, intervention status predicted neither the intercept nor linear slope (Bs = 0.17 and −0.08, SEs = 0.33 and 0.07, respectively, ns). AAs showed a significantly lower linear slope than EAs (B = −0.18, SE = 0.09, p < .05). Therefore, for every 1 year increase, the change of odds of AAs having any alcohol abuse relative to no abuse was 16% (1- e−0.18) less than that of EAs.
