Saturday, July 09, 2011

First Pop Ev Psych, now Pop Ev Sociology?

Remember those stories of a year or two ago about how things like obesity could be "socially contagious"?

Well, not so fast. It appears that they had a variety of statistical errors, the "research" behind them had never been published in a peer-reviewed scientific journal and other things.

Nicholas Christakis and James Fowler
That said, that didn't prevent the authors from giving a snazzy TED talk (it seems like TED is devolving more and more into pop science of various sorts and not always accurate pop science), and otherwise "selling" their findings.

That includes Fowler appearing on Colbert, where he made claims about losing weight himself as to not "infect" others. But, Fowler wants it both ways; he said he shouldn't have his research judged on comments like that.

Worse, showing once again that even actual science, not alleged science, isn’t perfect, the NEJM, per the story, is standing behind the claims.

And, that's sad, since the mainstream media never gives debunking stories the same play as it does original claims. If the NEJM isn't going to be a better gatekeeper, it's not so "good." Especially since it rejected a piece of debunking research. In short, professional journals like NEJM can "pull a Mooney," not just pseudoscience ones.

And, also "worse" is that Christakis and Fowler shoot legitimate research in the foot with antics like this:
So is obesity contagious? What about happiness and divorce and poor sleep? One irony of the contagion battles is that even if their methods are suspect Christakis and Fowler are obviously correct that peer influence exists and that it may be even more important than we realize. ...

But just because contagion is important in one context doesn't mean something like obesity spreads like a virus—much less one that can infect someone as remote from you as your son's best friend's mother. (For the record, I and my best friend's mother will eat our hats if it turns out to be true, as Christakis and Fowler claim, that loneliness is infectious, too.) Yes, we influence each other all the time, in how we talk and how we dress and what kinds of screwball videos we watch on the Internet. But careful studies of our social networks reveal what may be a more powerful and pervasive effect: We tend to form ties with the people who are most like us to begin with.
In other words, correlation is not always causation. And, to the degree causation is behind correlation, one had better get the correct order of cause and effect understood. Christakis and Fowler appear to be bad social scientists right there.

No comments: