In this paper, we develop an agent-based model of social influence on body weight. The model's assumptions are grounded in theory and evidence from physiology, social psychology, and behavioral science, and its outcomes are tested against longitudinal data from American youth. We discuss the implementation of the model, the insights it generates, and its implications for public health policy. By explicating a well-grounded dynamic mechanism, our analysis helps clarify important dependencies for both efforts to leverage social influence for obesity intervention and efforts to interpret clustering of BMI in networks.

The process of conditioning via reward learning is highly relevant to the study of food choice and obesity. Learning is itself shaped by environmental exposure, with the potential for such exposures to vary substantially across individuals and across place and time. In this paper, we use computational techniques to extend a well-validated standard model of reward learning, introducing both substantial heterogeneity and dynamic reward exposures. We then apply the extended model to a food choice context. The model produces a variety of individual behaviors and population-level patterns which are not evident from the traditional formulation, but which offer potential insights for understanding food reward learning and obesity. These include a “lock-in” effect, through which early exposure can strongly shape later reward valuation. We discuss potential implications of our results for the study and prevention of obesity, for the reward learning field, and for future experimental and computational work.

I develop a procedure for estimating local-area public opinion called stacked regression and poststratification (SRP), a generalization of classical multilevel regression and poststratification (MRP). This procedure employs a diverse ensemble of predictive models—including multilevel regression, LASSO, k-nearest neighbors, random forest, and gradient boosting—to improve the cross-validated fit of the first-stage predictions. In a Monte Carlo simulation, SRP significantly outperforms MRP when there are deep interactions in the data generating process, without requiring the researcher to specify a complex parametric model in advance. In an empirical application, I show that SRP produces superior local public opinion estimates on a broad range of issue areas, particularly when trained on large datasets.

Complex systems modeling can provide useful insights when designing and anticipating the impact of public health interventions. We developed an agent-based, or individual-based, computation model (ABM) to aid in evaluating and refining implementation of behavior change interventions designed to increase physical activity and healthy eating and reduce unnecessary weight gain among school-aged children. The potential benefits of applying an ABM approach include estimating outcomes despite data gaps, anticipating impact among different populations or scenarios, and exploring how to expand or modify an intervention. The practical challenges inherent in implementing such an approach include data resources, data availability, and the skills and knowledge of ABM among the public health obesity intervention community. The aim of this article was to provide a step-by-step guide on how to develop an ABM to evaluate multifaceted interventions on childhood obesity prevention in multiple settings. We used data from 2 obesity prevention initiatives and public-use resources. The details and goals of the interventions, overview of the model design process, and generalizability of this approach for future interventions is discussed.

IntroductionTobacco control policies focused on the retail environment have the potential to reduce tobacco use and tobacco-related health disparities through increasing direct and indirect costs. Recently, national and subnational governments have begun to restrict the sale of menthol products and reduce tobacco retailer density.MethodsWe developed an agent-based model to project the impact of menthol cigarette sales restrictions and retailer density reduction policies for six types of communities and three priority populations. During each simulated day, agents smoke cigarettes, travel in the community and make purchase decisions—whether, where and which product type to purchase—based on a combination of their own properties and the current retail environment.ResultsOf the policies tested, restricting all cigarette sales or menthol cigarette sales to tobacco specialty shops may have the largest effect on the total (direct and indirect) costs of purchasing cigarettes. Coupling one of these policies with one that establishes a minimum distance between tobacco retailers may enhance the impact. Combining these policies could also make the costs of acquiring cigarettes more equal across communities and populations.DiscussionOur simulations revealed the importance of context, for example, lower income communities in urban areas begin with higher retailer density and may need stronger policies to show impact, as well as the need to focus on differential effects for priority populations, for example, combinations of policies may equalise the average distance travelled to purchase. Adapting and combining policies could enhance the sustainability of policy effects and reduce tobacco use.