BS in Mathematics
The main objective of this paper is to analyze three different articles that discuss whether obesity could be socially contagious. According to the World Health Organization in 2013, obesity is the fifth leading risk for deaths around the world. This disease has dramatically increased in the last decade, which has led scientists to believe there are other factors contributing to the epidemic besides genetics. The first article I analyzed, written by Nicholas Christakis and James Fowler, provided a logistic regression model to estimate the odds of a person becoming obese. The model included the explanatory variables: age, sex, education, smoking behavior, geographical distance, social distance, and the BMI of a close friend. Christakis and Fowler found clustering of obesity in the network, and claimed it was caused by influence from one person to another. The second article, written by Ethan Cohen-Cole and Jason Fletcher, included environmental factors into the model and the coefficients for influence were no longer statistically significant. This led Cohen-Cole and Fletcher to conclude the clustering in the network could be partially explained by environmental factors. The last article, written by Cosma Shalizi and Andrew Thomas, claimed that when using observational data, it impossible to distinguish between mechanisms in networks. They provided a counterexample where they simulated a network with no influence and received results where influence was present in the model. This disproved the results made in the Christakis and Fowler article, claiming that influence causes clustering of obese people. Shalizi and Thomas provided the code they used in their paper. When I reproduced the results and changed the parameters, I found an example when the Christakis and Fowler argument may hold. Both of these networks were simulated, therefore more research needs to be done with a real network in order to see if the Christakis and Fowler claim is true or not.