By Michael Crupain, MD, MPH
The answer to that question, according to a well-publicized paper, published in the New England Journal of Medicine, is yes. The article published in 2007 by Christakis and Fowler used data from the Framingham Study (a long running cohort study from which we have learned a lot about chronic disease and its risk factors) to construct social networks to investigate the relationship between obesity and friendship. Using complex statistical methods, the authors argue that obesity is more likely to occur in a previously non-obese person if a close friend becomes obese and that this development of obesity is independent of their environment or a tendency of people to choose friends with similar attributes.
Today’s New York Times reports that another more recent paper is refuting the claims made by Christakis and Fowlers and states that the statistical reasoning they applied to their data does not allow them to draw the conclusions they have. In this new paper, author Prof Lyons, is quite annoyed by what he believes is the improper use of statistics (something that happens far too often) and he writes that in their New England Journal of Medicine Paper, Christakis and Fowler in many cases should have drawn conclusions opposite to what they did.