Here are some funny examples of the confusion between correlation and causation can go very, very wrong; courtesy of Tyler Vigan, the guy behind the blog. More, with 30,000 graphs; Spurious Correlations
More Skiing Leads to Increase in Bedsheet Entanglement Deaths: This chart shows a correlation between the number of people who die by becoming tangled in their bedsheets, and total revenue generated by U.S. skiing facilities. Clearly completely unrelated… but imagine the potential (inaccurate) headlines! Just think of the clicks!
Fewer New York Marriages Means Fewer People Murdered by Blunt Objects: Blunt object murders are going down, and so is the marriage rate in New York. These are totally unrelated facts, but when you put them together in a nifty little graph, it certainly can make you wonder if New York marriages are actually tragically deadly. But it shouldn’t! Because this is a spurious correlation, not a causation!
Rainfall in Tuscola County, Michigan, Leads to Murders by Pushing From a High Place: The more precipitation in Tuscola County, the more cases of people being murdered by being pushed from a high place. As you can see, they follow similar patterns.
If the study was done on humans, pay attention to who those humans were: The study is stronger if they’re randomly selected, nationally representative, and if there are a LOT of them.
Regarding research done on actual humans, who were the humans that participated in the study? Were they a huge group of randomly selected, nationally representative humans, or a small group that wasn’t randomly selected at all (say, 12 post-menopausal women with sleep issues, for instance)? The more people involved in the study the better. And the more that the people in the study represent the population as a whole, the more applicable the results are to the population as a whole.
Lots of psychology research, for instance, is conducted on college students, because they’re easily accessible on college campuses (where most research is being conducted). The problem is that college students aren’t a diverse group — they don’t represent the rest of the population in terms of age, race, socioeconomic status, you name it.
So those results might be fascinating, but you have to look at them critically — they’re really just saying that college students at this point in history at a mid-size university in the Midwest have these behavior patterns. It’s not saying that all people have these behavior patterns, or that they always have and always will.