This thesis investigates the interplay between adolescent friendship networks and behavioral tendencies over time, applying Stochastic Actor-Oriented Models (SAOM) and Separable Temporal Exponential Random Graph Models (STERGM) to longitudinal classroom data from a Dutch secondary school. SAOM is used to examine the co-evolution of network structure and behaviors such as delinquency, while STERGM is employed to disentangle the distinct processes of tie formation and dissolution. Across both frameworks, the results reveal robust structural patterns in adolescent friendships, with strong evidence of reciprocity and triadic closure. However, behavioral effects—such as homophily and peer influence on delinquency—were more variable and context-dependent, appearing more consistently in SAOM than in STERGM. SAOM captures feedback between individual behavior and network structure, while STERGM highlights the stability of core structural features and the contrasting mechanisms that govern the creation and dissolution of ties. Model fit assessments, including MCMC diagnostics and goodness-of-fit tests, confirm the reliability of both modeling approaches. These findings underscore the value of dynamic network models in understanding social and behavioral processes among adolescents, while also indicating the importance of larger datasets and longer observation periods for detecting behavioral influence and selection effects across diverse contexts.