It’s all connected! Except when it isn’t…correlation & causation


Do you get the joke in the above cartoon?

If not, read through to the end of the article and hopefully you will. And also understand one of the main mistakes we all make when learning from experience.

It’s often (rather smugly) said in science that correlation does not imply causation. This just means that just because two things happen at the same time doesn’t mean they’re connected. (Except, you know, in some broad sense where everything is connected…)

An alien arriving on earth might note that schools always seem to be full of children. Aha! Therefore the human race uses schools to reproduce!

In medieval times it was thought that lice were good for you. Sick people didn’t seem to have any. Therefore the lice kept you well. It turns out, however, that lice are sensitive to body temperature and so they tend to leave when someone has a fever.

correlation and causation
One of Tyler Vigen’s spurious correlations www.tylervigen.com

Tyler Vigen specialises in meaningless examples of when two things seem to happen at the same time. Here we can see that how much cheese is consumed in the US seems to be connected to how many people die from getting tangled up in bed sheets. I never knew that was a thing.

Ok, but why does this matter?

When I was younger the schools always tried to instill in us the idea that marijuana was a gateway drug. Smoke some dope and there was a strong likelihood you’d end up as a heroin addict. After all, almost everyone who took heroin had started with marijuana.

It made sense! Sort of. Except that it didn’t really.

All those heroin addicts probably also tried coffee at some point. And cookies. But no one was blaming the coffee or the cookies for the heroin addiction.

When two things happen together it makes for a strong story, however. It seems that humans just love to connect the dots.

Take the anti-vaccination campaigners who give examples of where a child developed autism after receiving a vaccine. Watching your child develop autism must be absolutely awful. I can’t imagine the anguish. And it’s natural to want to point the finger at a single cause.

There is no evidence that vaccines cause autism but as most young children are vaccinated and autism usually develops in young children, there is a correlation. Some other things that might happen at the same time might include:
– switching from mother’s milk to cow milk
– watching movies for the first time
– the mother getting pregnant again

In fact, before the vaccine/autism theory came along parents of children with autism used to blame particular life events that might have caused it.

So when two things happen together there’s no connection?

Well there might be. When two things seem to happen together a lot it could at least mean that it’s worth further investigation. If I wake up with a headache each time I drink whisky the night before I might try giving it up for a while to see if it makes a difference.

But it’s frequently an illusion that’s used for political gain.

Suppose you come across a social media post that shows how crime goes up when immigration goes up. The clear conclusion: immigrants are criminals!

But maybe the crime rate was going up for years before immigration increased. Maybe there are other factors like an economic crisis that causes both a crime wave and more immigration.

Here’s the cartoon again. Does it make sense now?

(he’s casting doubt on the connection between him taking the statistics course and understanding the relationship between correlation and causation. A joke for nerds.)

 

Further reading:

https://www.iperceptions.com/blog/causation-vs-correlation – more on correlation and causation with some good examples

https://www.tylervigen.com/spurious-correlations – more silly graphs from Tyler Vigen