It’s just about Halloween and shops are full of the scary and spooky costumes and decorations. But the scientists have had some fun too.
Inferring causation from correlation
You may not think that this is actually something scary, or maybe you have no idea what it means. I hope to explain what this means and why it is a legitimate issue.
Causation is defined by the Oxford English Dictionary as “the action of causing something” and correlation is defined as “a mutual relationship or connection between two or more things”.
A common problem when science is communicated in news articles is confusing these two very seperate concepts to get headlines like “Coffee Causes Heart Disease“, “Reading National Geographic makes you smarter” or “Margarine consumption linked to divorce”. They’re great flashy headlines and very good at grabbing your attention when you’re scrolling mindlessly through your Facebook feed.
The problem is that most of these are actually based on studies that show that many of the people who drink coffee do suffer some form of heart issue, or that many intelligent people do read National Geographic and that both margarine consumption and divorce rates have been falling in recent history. But what if it’s not actually coffee that is the root cause of heart disease and instead it’s the lifestyle adopted by many coffee drinkers? Perhaps people of lower intelligence are less likely to be interested in reading National Geographic. And it’s great that divorce rates are falling, but it’s unlikely to be the result of some ‘mysterious chemical’ in margarine and far more likely to be that people aren’t eating as much margarine any more.
Some more hilarious examples illustrate this point below:
These examples all show quite unrelated data and it’s very easy to question the link. Just because the two are shown on the same graph does not mean that one thing causes another to occur- just that they happened concurrently.
However, it’s harder to apply the same principles to scientific news reports because most people have no idea whether some of the chemicals in coffee might damage the heart. In fact, many people would recognise that caffeine affects the heart and would accept this claim without thinking twice. And all of a sudden, many people believe they should avoid coffee at all costs. To actually make this claim, the original scientific paper would have to have identified the mechanism for coffee causing heart disease rather than a casual correlation between the group of people who like coffee and who also have heart disease.
And this misinterpretation of the two concepts, correlation and causation, can be quite scary when a news headlines goes viral without actual data and mechanisms to support it.
Related is also the survivorship basis: did you know that cats surviving falls from floors below the sixth tend to get more serious injuries than the ones falling from higher floors?
It’s just a way of manipulating statistics without an underlying understanding and it’s not helpful or useful (unless you happened to be running a company that sold tea, I guess….). So please remember a very simple phrase this Halloween, correlation does not imply causation!