Tuesday, December 09, 2008

Correlation v. Causation

Remember how a few days ago I posted about interesting articles in the NYT?

One of the articles that I noticed was "Strangers May Cheer You Up, Study Says,"which was a new study which had results showing that like, friends of friends could cause you to be happy. I mentioned a bit about it, and said that it was interesting.

On the Freakonomics blog (which, incidentally, is a NYT blog...), today there's an article saying the study is bunk. The Freakonomics article is more interesting than the first article that I mentioned, actually.

It says that the study was performed sloppily. The study proves correlation but not causation. This article says that there are actually three reasons that you may be happy when strangers near you are happy-- the first is that happiness actually is contagious. The second is that "people with similar dispositions are more likely to be friends," and the third (and Justin Wolfers says this is most likely) is that people who are friends are more likely to be under similar influences. The example he gives is, if our mutual friend dies, we'll both be sad about it.

Wolfers also mentions another article in the same issue of BMJ which proved that people can too easily draw conclusions about contagion.

"They use Fowler and Christakis’s approach on another dataset, and show that it leads to the unlikely conclusion that height, headaches, and acne are also contagious. The more likely explanation, of course, is that all are subject to similar environmental influences. For instance, the same jackhammer causing your headache is likely causing mine."

I remember when we talked about correlations in my psychology class. Our teacher said, "Ice cream sales are strongly correlated with crime rates. Therefore, ice cream causes crime." And then we talked about it, and of course that isn't actually true; when it's hot, people buy ice cream. And there is also more crime when it is hot, probably because people are more irritable.

No comments: