2 comments

  1. Lynde
    My statistics professor repeated so many times that if we remember only one thing from his class, it should be that correlation does not imply causation. He wrote it on the board at the beginning of every class for a couple of weeks, and I think he even made us repeat it out loud a few times. Reading the news, not only on this subject but also on social science, I see what made him speak so passionately about this. I think of him every time I read one of these stories. Thanks, Dr. Bernstein!
  2. I think what he meant is that association does not necessarily imply causality. Strictly speaking, nothing implies causality and all we have is association. The question is whether the association is strong and whether it has an underlying mechanism. I raised the question in my recent blogpost http://wp.me/16vK0 where I gave an analysis of an important experiment from Ron Krauss's lab studying both carbohydrate and calories. After our analysis was published, Mike Eades suggested that the controlling variable (rather than how long you had been on a particular diet) was the amount of carbohydrate. When we plotted out the data there was a very strong association between carbohydrate and, in the case I show, triglycerides. It is because the association was strong and because we know something about mechanism. It's good to consider too that the case is stronger because it was turned into an observational experiment (that is the authors had not set out to look at carbohydrates but also at calories as a controlling variable so the results have no bias). Association does not necessarily imply causality but sometimes it's really strong.

Leave a reply

Reply to comment #0 by

Older posts