There's been a rash of studies in the news purporting to show that some medication, food, or behavior (call it X) has an effect on some aspect of health (call it Y). In Psych 42, I've talked recently about new studies on caffeine and miscarriage and anger suppression and mortality (i.e., death). The latest study concerns birth control pills and ovarian cancer.
All of these studies do clearly show an association between X and Y. However, because none of them is an experiment--where the independent variable is manipulated and extraneous variables are controlled--they do not show that X causes Y. And it doesn't matter how many times the journalists or researchers state that they do.
All of these studies do clearly show an association between X and Y. However, because none of them is an experiment--where the independent variable is manipulated and extraneous variables are controlled--they do not show that X causes Y. And it doesn't matter how many times the journalists or researchers state that they do.
In the latest example, the researchers showed that women who took the Pill were less likely than women who did not take the pill to develop ovarian cancer. This could be because they took the Pill ... but it is also possible that women who took the Pill differed from women who did not in some other way. For example, women who took the Pill might also have been women who tended to think about and act on their health and well-being more than women who did not--and this difference might have been responsible for the reduction in ovarian cancer.
Yet the article linked above is titled "Pill Prevents Ovarian Cancer for Decades." The word "prevents" clearly implies causation. But if the risk reduction the researchers found was due to something other than the Pill, then then it is not the Pill that is doing the "preventing" and going on the Pill will not reduce anyone's risk.
There are other issues here--such as the effectiveness of statistically controlling for confounding variables--but I'll cover those in another blog. For now, let me give my standard disclaimer about this sort of thing. I am not saying there is anything wrong with this research. It is interesting and important and tells us that there is an association between X and Y that might be a causal one. But researchers, journalists, and the general public need to be much more cautious about jumping to causal conclusions.
Yet the article linked above is titled "Pill Prevents Ovarian Cancer for Decades." The word "prevents" clearly implies causation. But if the risk reduction the researchers found was due to something other than the Pill, then then it is not the Pill that is doing the "preventing" and going on the Pill will not reduce anyone's risk.
There are other issues here--such as the effectiveness of statistically controlling for confounding variables--but I'll cover those in another blog. For now, let me give my standard disclaimer about this sort of thing. I am not saying there is anything wrong with this research. It is interesting and important and tells us that there is an association between X and Y that might be a causal one. But researchers, journalists, and the general public need to be much more cautious about jumping to causal conclusions.
1 comment:
I agreeed with that!
That's how the stereotype start!
Post a Comment