I found a great little infographic on correlation vs. causation, which is one of the most important lessons every science-minded person must learn in order to effectively understand the world. People have the tendency to make assumptions based on things they don’t understand, because they have not learnt the fundamentals of critical thinking. For example, many people buy cold medicines, which end up “completely curing” clients after a week of usage. Most people would chalk their recovery up to the medicine, when it really may have been nothing more than just a function of time. So this article will clear up some misconceptions involved in the correlation/causation confusion.
Explaining the Variables
When it comes to dating, or even making a first impression (like in the picture above), people always debate about “what really matters.” For example, good looks, money, a buff physique, etc. Many people (erroneously) believe that money is extremely important, which is to say… “wealthy men have great success with women.”
If you surveyed 1000 men and found that the best looking ones have the most success with women, it would be reasonable to infer that there is a relationship between the two variables (as opposed to, for example, if we only asked 3 people). As a formula, we could say “money,” or variable X, causes “success with women,” or variable Y (X->Y). But this may not necessarily be the case. We can reasonably assume that success with women doesn’t magically turn a man into a good looking guy (i.e., Y doesn’t cause X), but there is another explanation still. A new variable, such as “confidence,” or variable Z, may in fact cause both of these things. A confident man gets promoted more than a shy one, which explains the wealth, and women are more attracted to confidence than shyness, which explains the success.
Every student of psychology learns about these (x, y, z) variables early on, and I have a million random examples that I could use. But Seomoz.org has an interesting graphic that goes into greater depth for types of relationships between variables, so allow this to educate you:
“Correlation is Not Causation!”… Or is it?
I often catch people making the mistake of saying “correlation is not causation,” which always makes me cringe a little. Okay, true, correlation is not causation… when it’s not causation. But by definition, causation must be a correlation. If something causes something else, then there’s obviously a relation there. If you ever get confused about what to say, just re-read the title of this post.
There are some instances, also, where we basically know that a correlation implies a causation, as mentioned in the graphic above. For an additional example, when the rates of car crashes on a highway skyrocket after it becomes very icy on the road; we know the car crashes didn’t cause the ice to form on the road, so it’s probably the other way around.
You could argue that, yes, perhaps a third variable – like aliens – caused the ice to form and the cars to crash, but critical thinking isn’t just for ivory-tower academics, so let’s stay realistic. The icy roads probably caused the increase in car crashes. In fact, it’s this ability to make inferences about variables that make people able to “conduct experiments” within the “science of history.”
History sometimes has “natural experiments,” in which, often by random virtue of having numbers readily available (due to public records, for example), people can make inferences that inform us about our past. This is what esteemed geography and physiology professor Jared Diamond at UCLA did in his famous book “Guns, Germs, and Steel.” He argued, for example, that geography – and, by extension, food production – was a huge factor in the success of societies, using knowledge from earlier records.
Here’s a brief video that basically talks about the idea of a “lucky shirt.” In terms familiar in this article, they basically explain that just because a basketball player started performing well after getting a new shirt, does not mean that the shirt caused his performance to improve.
The Real Message
Most of the errors I see are in the places most people see them – news stories. People will misinterpret or mislead with information in order to sell papers and grab viewers’ attention. So keep your eyes open for those headlines that say things like “Study: Handsome men wear bigger shoes” or “Scientists prove that confident women are bitchy.” And even more importantly, read past the headlines (while reminding yourself yet again that “correlation does not imply causation”).