Network Theory in Health Metrics

The heavy-tailed/small-worlds crowd had a big impact in health research recently, and now it’s drawing criticism from the theorists. Slate covered the story well a couple weeks ago, and interviewed Russ Lyons at length about the methodological shortcomings of the evidence that obesity, smoking, and loneliness are socially contagious. The Slate article even links to Russ’s preprint on the matter, which is some pretty technical stuff to point a general audience towards. Go Slate!

When I passed on some of Russ’s concerns to my experienced health metrics colleagues, one replied that the idea of social contagion is important enough that it doesn’t matter if the methods are wrong. Interesting perspective. It reminds me of Gian Carlo Rota’s ordering of mathematical results: most important are definitions, less important that that are theorems, and much, much less important than that are the proofs.

I’ve been in meetings for almost 3 weeks now, and meanwhile more good papers on networks for health are pouring out. Christakis and Fowler have posted a preprint to arxiv, showing how network thinking improved flu surveillance of Harvard undergrads. So maybe the idea was the important part. Meanwhile, the Cosma Shalizi and Andrew Thomas have an additional critique preprint, to be put in the same category as Russ’s. I asked Russ what he would accept as evidence of social contagion, and I didn’t find out, but the paper by Shalizi and Thomas says maybe nothing can be convincing: Homophily and Contagion Are Generically Confounded in Observational Social Network Studies.

For me, it’s time to get back to that meeting!


Filed under global health, statistics

2 responses to “Network Theory in Health Metrics

  1. Pingback: Societally-useful applications of Networks « Healthy Algorithms

  2. Pingback: Network Stats Continue « Healthy Algorithms