Social Networks (in the wild)
I attended the New England Statistics Symposium (NESS) last Saturday, and I’ve been meaning to write about one of talks I saw. After lunch, I went to the Columbia section so I could see the talk about multiple imputation using chain equations. The MI talk was the second in the section, so I sat through the first talk presented by Tian Zheng (Tian’s Blog) which turned out to be very interesting. The talk was about using social networks to learn about at risk populations.
My understanding of this is that a survey could be given asking people questions about who they know rather than about themselves. For instance, instead of asking “Is your name Michael?” and “Are you homeless?” ask “How many Michaels do you know?” or “How many homeless people do you know?” Then using the responses to these questions, researchers can estimate how large at risk populations are. And they can do this without ever asking people who are in the at risk population! Really neat.
Why is this useful?
This excerpt from this flyer that was created to describe the method to a general audience says it very well:
“AT-RISK POPULATIONS: At-risk populations can be hard to access (eg. homeless) or reluctant to admit their status for fear of others finding out (eg. HIV/AIDS, drug abusers, sex workers). Statisticians learn about these populations through their friends and acquaintances. Instead of asking if a person uses IV drugs, ask ‘How many IV drug users do you know?’ and use social structure to learn about the person using IV drugs.”
Really, really neat stuff.