|
|
Skirts don't Cause CancerCorrelation is not causation
|
|
|
|
|
This principle, skirts don't cause cancer, is an easy way to show people that correlation, weak or strong, doesn't necessarily cause something that it appears to cause. This principle shows that reality could be the opposite of what people see and think it is.
Whenever you read that a study shows that X "causes" Y, you have to ask yourself if the study really shows that X is a cause of Y, or if there is only a correlation between the two. Studies have a big problem in showing that X causes Y because the quality that's required to show this causation costs a lot of time and money. Instead, studies usually limit themselves to showing only that X and Y are correlated, that is, that they are linked and move together when one of them moves. Correlations, especially weak ones, can't prove anything because reality is not a perfect distribution of probabilities, and there's always a small bias due to various local factors. Believing that a correlation proves something is a form of observational bias. You could find (weak) correlations in virtually anything. For example, you could find that high cancer percentages are correlated with people sleeping (or not) in pajamas, or with their socks on or off, or with their heads toward a specific cardinal direction, or wear (or not) slippers in their homes. In fact, you could find a strong correlation between high cancer percentages and people who wear skirts. Let's say that a "scientist" conducts a study to see if wearing skirts, even if just a few times in life, causes cancer. He concludes that people who wear skirts (instead of pants) are more likely to have breast cancer. Then, all sorts of people on the Internet say that the study proves that wearing skirts causes cancer. It's actually true that wearing skirts is strongly correlated with breast cancer. If you don't look at all the factors (presuming that you can find them), if you believe in correlations, you end up believing that wearing a skirt causes cancer. Some people will realize that another factor is responsible for the higher cancer percentage: being a woman. You know about this factor because you've heard before that breast cancer affects many women (but very few men), and that women's biology leads to an increased probability of breast cancer. And even if you didn't hear it, you could easily see this factor because you can easily distinguish between men and women, so you can easily make a logical connection; children have to be told about it. This reality is that skirts don't cause cancer, but they may be correlated with certain types. What if the same "scientist" conducts a study to see if wearing skirts or pants affects the percentages of cervical cancer in women? The result could show that wearing skirts correlates with a higher percentage of cervical cancer, but it could also show that wearing pants correlates with a higher percentage of cervical cancer. For the "scientist" it's irrelevant what the correlation is, and publishes a study titled "Skirts / pants cause cancer". The reality is that either of them, skirts or pants, may be correlated (weakly or strongly) with a life style that's slightly more prone to cause cervical cancer, so this life style may be the cause of the cancer. The reality is that neither skirts nor pants cause cancer, and why correlations, weak or strong, don't show causation. |
|
|
|
|
|
|
|