Seminar SERIES

Department of Epidemiology, Biostatistics, & Occupational Health
McGill University, Montreal, Quebec

“How to Break Something By Fixing It”.

Recent years have witnessed a tremendous disciplinary cross-pollination in which epidemiology has borrowed heavily from other methodological traditions, for example the infusion of econometric methods into social and perinatal epidemiology. Methodological traditions arise within a field from its unique history and culture, including philosophical approaches to quantification and causal inference. To exemplify some of the differences in worldview between epidemiology and economics, I will highlight the handling of clustering, as both an analytic challenge (hierarchical data) and as a design strategy (matching). Causal tools from one discipline reveal the limitations and dangers of the practices that characterize the other. I review the good, the bad and the ugly from fixed effects designs like exposure-discordant siblings, and how insights from both epidemiology and economics balance the pros and cons of this method in health research.