Alcohol use and misuse are key contributors to worldwide burden of disease, and they are leading causes of mortality in adolescents and young adults. Alcohol use is commonly initiated in early-to-mid adolescence and grows strongly in subsequent years. Much previous research on adolescent drinking has treated adolescents as a homogeneous population, and has focused on identifying general protective or risk factors that contribute to alcohol use. However, recent studies have shown that while most adolescents show slow and steady increases in alcohol use as they get older, some adolescents show atypical and rapid escalation in alcohol use. Trajectory-based approaches have been used in the empirical literature, but there are questions about the robustness of these techniques when assumptions are violated, which is commonly the case. It is also likely that the potency of various social influences varies across ages and key transitions such as transition from primary to secondary school. This thesis first examines the robustness of a type of trajectory-based model, Latent Class Growth Analysis, in identifying different trajectories in a population. After establishing the appropriateness of this technique using progressive simulations based on real-world distributional and sample variability (paper under review), the thesis examines potentially time varying social influences, including parents, siblings, peers and school factors that contribute to rapid escalation (paper in press – Addiction). The main analytical sample is based on seven waves of data (collected from 2002- 2009) from 927 children (Grade 5 at entry). Lastly, this thesis examines the utility of a relatively novel analytical technique, Cusp Catastrophe Modeling (CCM), in facilitating our understanding of rapid escalation. The thesis outlines a new evidence base for time varying social influences, and provides guidelines for the appropriate use of high level statistical modeling of this type of data.