This dissertation studies the social structures and dynamics of human networks: how peers at the micro level and physical environments at the macro level interact with the individual preferences and attributes and shape social dynamics. It is composed of three parts. The first essay, "Friendship Choices and Group Effects in Adolescent Smoking" explores the Add Health network data for modelling of peer effects. It analyzes the association between group effects and individual behavior, as well as how the composition of friendship choices is affected by the change of an individual's attribute. This paper acts as exploratory analysis and theory building piece for the second paper. The second essay, "Social Distance and Homophily in Adolescent Smoking," addresses the issue of peer selection vs. peer influence. Human social networks are characterized by high levels of homogeneity and clustering, and the question it seeks to answer with the study of adolescent networks is which of the two dynamics is most responsible for the problem of adolescent smoking. It employs the concept of social distance to parse out the effects of selection and influence: the key insight is that influence and selection, while seemingly confounded, are differentiable with the use of social distance. Friendship between two peers socially distant implies strong selection effect: the effect of influence becomes weaker as distance grows. It also addresses the concern for selection on observables by adjusting the findings with the propensity score weights model with three treatment indicators. In the third essay, "Collective Location for Collective Action," the paper discusses collective action and collective location problems as complex social dynamics that are shaped by physical factors. Collective action can be seen as an example of multiple prisoner's dilemma game, in which the Pareto inferior Nash equilibrium is to always defect. Mass protests are collective actions taking place in a single location. The provision of collective action then depends on the solution to the collective location problem. It shows that by solving the n-person location problem, the solution of which is the center of mass of any convex surface, mutual cooperation becomes the Nash equilibrium solution to the collective action problem. Three appendices are included: (1) Plots of School Trends in Initiation and Cessation; (2) Detailed Regression Results (Second Essay); and (3) Analysis of the Korean Protest Data. Individual sections contain references. (Contains 20 figures, 13 tables and 7 footnotes.) [This document was submitted as a dissertation in May 2010 in partial fulfillment of the requirements of the doctoral degree in public policy analysis at the Pardee RAND Graduate School.]