Clustering and spreading of behavior and opinion in social networks Lazaros Gallos Levich Institute, City College of New York Hernan A. Makse - Shlomo Havlin
Clustering and spreading of behavior in social networks Lazaros Gallos Levich Institute, City College of New York Hernan A. Makse - Shlomo Havlin
Obesity epidemic (?)
BMI and obesity The Body Mass Index (BMI) is a standard measure of human body fat BMI>30 is generally accepted as the obesity threshold
Obesity in USA increases with time
What we know on obesity ‘spreading’ 1.Genetics 2.Peer pressure (Christakis and Fowler, NEJM, 2007) 3.Spatial clustering
Our approach The physics of clustering is challenging Study obesity as a percolation process Use scaling analysis More properties
Obesity prevalence in USA
Percolation transition
Time evolution of obesity clusters County obesity %
Largest clusters County obesity %
Neighbors influence (after Christakis, Fowler)
Distance-based correlations
The increase rate is also correlated
Spatial correlations: Scaling theory of Growth Standard theory of Gibrat assumes random growth Scaling concepts introduced by the H.E. Stanley group (Stanley, Nature, 1996) for the growth of companies Extended to more properties (e.g. cities) Growth rate:
Limits High correlations: No correlations: =0, =0 =0.5, =2 (in 2d)
Spatial correlations (constant in time) =0.5 Obesity =1.0 Population
Digestive cancer mortality (Changes with time)
Time evolution of
Phase diagram Uncorrelated Random walk Human activity Economy City growth Population Mortality Cancer mortality Obesity Diabetes Inactivity Lung cancer / d
Conclusions Strong spatial correlations in obesity spreading Obesity clusters grow faster than the population growth Scaling analysis quantifies the degree of spatial correlations Exponents are related Three main universality classes based on spatial correlations