Spread of Acne, Headaches, and Height in Social Networks?

One of my collaborators, James Fowler, has been in the news lately (yet again!) for his work on the Framington Heart Network. This time, he and his collaborator Nicholas Christakis (who is an excellent speaker and a nice guy, by the way) have looked at the spread of happiness in social networks. They previously published papers about the spread of smoking (and non-smoking) and obesity in these networks. (I was thanked in the obesity paper, as I had read and commented on a draft of it a year before the final version came out. When Nicholas gave this talk, I loved how the obesity level of a person was indicated by the radius of the corresponding node.)

One thing I just found out by taking a look at improbable.com is that another relevant study, that gives a warning about the dangers of overinterpretation, appears in the same issue of the British Medical Journal. As improbable.com states, it's highly amusing that the journal issue includes [a] strange juxtaposition of a study and a study mocking it. You can read the BMJ's editorial here.

The social networks literature from the social science side of things is extremely interesting, but one thing that I do see in a lot of that research is a tendency towards overinterpretation. My side of the scientific literature has its own set of faults, and in all of these cases, it's good to be reminded about these dangers now and then. We all need to make sure to keep our feet on the ground. (By the way, I am in no way denigrating the study. I think that it's a very nice piece of social network analysis. I'm just pointing out the need for vigilance and the excellent point also made by the other paper.)

I couldn't find a link to get this individual blog entry from improbable.com, so I have copied the text below:

Much excitement in the news about a study just published in BMJ (British Medical Journal):

“Dynamic spread of happiness in a large social network: longitudinal analysis over 20 years in the Framingham Heart Study,” James H Fowler and Nicholas A Christakis, BMJ 2008 337: a2338. The conclusion: “Peoples happiness depends on the happiness of others with whom they are connected.”

Which, of course, is a causal conclusion.

But lesser attention appears to have been paid to another study published simultaneously in the same issue: “Detecting implausible social network effects in acne, height, and headaches: longitudinal analysis,” Ethan Cohen-Cole and Jason M Fletcher, BMJ 2008;337:a2533

They found that a friend´s acne problems increased one’s own acne problems, a friend’s headaches increased one’s own headaches, and a friend’s height increased one’s own height. Given the first two, it seems one is better off without friends.

Their conclusion: “Researchers should be cautious in attributing correlations in health outcomes of close friends to social network effects, especially when environmental confounders are not adequately controlled for in the analysis”

Now see the first study again.

The same issue of BMJ also contains an editorial that gives further insight into this strange juxtaposition of a study and a study mocking it.