Bayesian graphical modeling and its application to urban mental health

Mental health problems, such as depression, alcohol use disorder, and anxiety, are increasingly prevalent and contribute substantially to the global disease burden. With over 50% of the world's population living in cities, expected to rise to 70% by 2050, researchers are exploring the unique urban factors contributing to mental health issues. The urban environment impacts mental health through several mechanisms, such as differing sociodemographics, limited access to resources, and lack of green spaces. The network approach to psychology has been proposed to understand these complexities; it views mental disorders as arising from a set of interconnected affective, cognitive, and behavioral variables, offering a new theoretical framework for their study. However, this approach requires estimating many parameters, casting doubt on whether common sample sizes are sufficient to derive robust inferences from the networks. Bayesian graphical modeling has been proposed as the alternative to the common, frequentist network estimation and can quantify the statistical evidence supporting one's networks. This dissertation addresses the questions: How can we optimally assess and model the interplay of variables influencing urban mental health? Additionally, what methodological advancements are essential for an accurate and comprehensive evaluation of these factors? The dissertation invites readers to reconsider common network analysis approaches, question the statistical evidence underlying one's network, dive into Bayesian alternatives to common estimation approaches, and use these statistical advances to understand key (urban) mental health challenges.