Wednesday, April 22, 2009

Correlation does not imply causation

In today's New York Times, Tara Parker-Pope writes about the benefits of friendships when it comes to an individual's long-term health. She says, 
In 2006, a study of nearly 3,000 nurses with breast cancer found that women without close friends were four times as likely to die from the disease as women with 10 or more friends. And notably, proximity and the amount of contact with a friend wasn’t associated with survival. Just having friends was protective.

There are a number of statistical relationships that she cites between friendships and survival. In fact, a stronger relationship exists between health and having a support system of friends than health and family. I don't doubt that there's some truth. But I also think back to my Research and Statistics course where it was drummed into us: Correlation does not equal causation. 

As soon as a statistical link is demonstrated, we have to look at the confounds, the alternative explanations. If someone has a strong network of friends, non-related support, then that tells us about that person. The same things that cause an individual to have many friends may be the same things that cause that person to fight cancer.

No comments: