Behavior, attitudes, and infection can transmit across networks of contacts via social mixing, making network analysis methods a key tool in social and infectious disease epidemiology. Through analysis of the simultaneous processes that influence and shape individuals and networks, we can better understand how to collect social network data, incorporate human behavior and its collective idiosyncrasies into models and statistics as uncertainty, and thus improve the veracity of our conclusions. Using data from a longitudinal social network study of undergraduate students, this dissertation aims to: 1) examine how social structures and contact patterns shape alcohol consumption and use in undergraduate students; 2) evaluate the strengths and limitations of different methods of measuring social contact networks; and 3) develop methods to quantify network uncertainty and hypothesis testing for trait assortativity. First, we applied social network analysis methods to two undergraduate student social networks, investigating network correlates of alcohol consumption, identifying numerous, consistent associations between alcohol use and social position in this population. Specifically, network position, alcohol exposures, and relationship strength were associated with individual alcohol use, suggesting complex relationships between drinking and network topology, as well as proximity to alcohol use. Overall, this chapter adds to the body of evidence of significant relationships between network structure, social position, and alcohol consumption. Next, we systematically compared two social network measurement methods with varying levels of granularity in order understand the unique utility of self-report vs. sensor contact data, as well as trends in data quality and quantity over time. Networks were compared across and within each measurement method, using overall network structure, dyad, and node characteristics. We found few network similarities between measurement methods, suggesting that neither empirical network measurement method are complete representations of the underlying “true” social network. These analyses highlight the impact that network measurement can have on empirical network findings and suggest that researchers should carefully consider which collection method, or combination of methods, could provide them with the highest quality data needed to answer their research questions. Finally, we outlined and defined multiple assortativity sensitivity analyses, uncertainty quantification approaches, and null model-hypothesis testing procedures and applied these methods to a measured social network of undergraduate students. These investigations showed that uncertainty and biases of attribute assortativity may be predictable, given a defined amount and type of data error. Generally, results of these analyses show the potential impacts that data quality, measurement error, and the measured network can have on observed assortativity. We suggest that it be standard practice to conduct and present assortativity sensitivity analyses, and to hypothesize possible confounding or bias related to network data quality and completeness. In toto, this dissertation describes and extensively explores social networks of undergraduate students. We investigated relationships between a risky health behavior of public health importance and network features, as well as how network analysis results using observed networks are reliant on the network measurement method and the types and amounts of data uncertainty and error present. These projects have generated new results and insights into alcohol use and social networks in a college setting, compared empirical social network observations between a traditional and novel instrument, and developed a suite of analytical social network tools. Importantly, the novel methods we defined and implemented in this dissertation provide a framework with which to evaluate network uncertainty, robustness, and hypotheses.