Sensitivity analysis for contagion effects in social networks.

Area of Science:

• Social network analysis
• Epidemiology
• Behavioral science

Background:

• Social network analyses suggest behaviors like obesity and smoking spread through social ties.
• These findings face critique due to potential homophily (similarity) and confounding factors.
• Similar patterns observed for non-contagious conditions like headaches raise doubts.

Purpose of the Study:

• To assess the robustness of social contagion effects for obesity, smoking, happiness, and loneliness.
• To determine if homophily or unmeasured confounding explains observed social network associations.
• To evaluate the relevance of critiques using data on height, acne, and headaches.

Main Methods:

• Utilized sensitivity analysis techniques on social network data.
• Applied methodology to both health behaviors (obesity, smoking) and subjective states (happiness, loneliness).
• Compared findings with data from traits like height, acne, and headaches to test critique validity.

Main Results:

• Contagion effects for obesity and smoking cessation showed robustness against homophily and confounding.
• Contagion effects for happiness and loneliness were less robust and more susceptible to explanation by homophily.
• Observed associations for height, acne, and headaches were easily explained away by homophily and confounding.

Conclusions:

• Social contagion effects for obesity and smoking cessation appear reasonably reliable.
• Observed associations for happiness and loneliness require cautious interpretation due to potential confounding.
• Sensitivity analysis is a valuable tool for distinguishing true social influence from homophily in social networks.