Table 4 summarizes parameter estimates from the multilevel growth model where we disentangled between-person and within-person variations and predicted frequency of alcohol consumption from positive/negative reinforcement, AD, and the two 2-way interactions between them. Random slopes were retained for linear and quadratic effects of age, AD, and the interaction effect between negative reinforcement and AD because preliminary analysis indicated that they were statistically significant. Results indicated that, at the between-person level, positive reinforcement was associated with more frequent alcohol consumption, but negative reinforcement was not associated. At the within-person level, positive reinforcement was also associated with more frequent alcohol consumption. Contrary to our hypothesis, this association did not vary as a function of AD, as indicated by the non-significant interaction effect between positive reinforcement and AD. Consistent with our hypothesis, there was a significant interaction effect between negative reinforcement and AD. Specifically, the parameter estimate of the main effect of negative reinforcement indicated a positive association with alcohol consumption when AD diagnosis was absent. The positive parameter for the interaction between negative reinforcement and AD indicated that the within-person association
