Abstract: Sociologists, economists, epidemiologists and others recognize the importance of social networks in the transmission of ideas, behaviors, and health outcomes through human societies. Contagion is a central concept in research on diffusion, cascades, peer effects, spillover, and interference in social, communication, and epidemiological networks.  Researchers have offered varying formal definitions of contagion, and there is disagreement about whether popular statistical approaches can detect it in real networks. Learning about the effect of individual-level treatments (e.g. vaccination, education, incentives) under contagion is difficult, with researchers offering competing claims about the validity of causal effect estimates. In this talk, I will present a unified view of contagion in networks, with emphasis on empirical challenges faced by researchers working with network data. I first review the conceptual approaches to defining contagion, and explain how contagion differs from dependence, clustering, and “treatment spillover”.  I will present an explicitly causal definition of contagion and explain its temporal mediation structure. I show that some regression approaches — in particular those often thought to be “robust to dependence” — can deliver profoundly inaccurate estimates of individualistic treatment effects under contagion.  I outline new causal intervention effects under contagion, and show formally why randomization may not be sufficient to eliminate confounding under contagion.  I give empirical examples from studies of infectious disease transmission and adoption of health-related interventions in a randomized trial.

 

Bio: Forrest W. Crawford PhD is Associate Professor, Department of Biostatistics, Yale School of Public Health, Yale School of Management

(Operations), and Department of Ecology & Evolutionary Biology, Yale University. He is affiliated with the Center for Interdisciplinary Research on AIDS, the Institute for Network Science, and the Computational Biology and Bioinformatics program. He is the recipient of the NIH Director’s New Innovator Award and a Yale Center for Clinical Investigation Scholar Award. His research interests include causal inference, networks, graphs, stochastic processes, and optimization for applications in epidemiology, public health, and social science.