Correlation and causation can be tricky to grasp – it’s not always clear how they fit together. Just because two things are linked doesn’t mean that one causes the other… except when sometimes it does.
The basics of this rather complicated relationship are put very well in the video below, from MinutePhysics. It uses the example of a (fictional) correlation between owning a cat and being above average height, so we’ll follow the same example to be as clear as we can.
If you’ve got statistics that show cat owners are generally taller, that’s a correlation, not a causation: those figures on their own can’t prove that having cats around the house is making people taller, or that taller people have a deeper yearning to look after cats.
Correlations can be suggestive though, especially if a lot of data is involved – in our example, while we can’t prove cats help their owners to grow an extra few centimetres, it does look like something is going on beyond random chance.
The question is, what?
Scientists will often make suggestions about how one thing could influence another, and that’s usually helpful – these ideas can then be put to the test in other studies. Just make sure you remember the difference between statistics and hypotheses clear in your mind.
As MinutePhysics points out though, correlation can imply causation, if we’ve got a broad enough set of statistics to go off, thanks to causal networks. Simply put, that means more data on more contributing factors.
An example would be knowing the height of the cat owners before they had a cat – if the heights went up after the cats moved in, the hypothesis that the cats were the cause seems a reasonable idea to investigate.
Instead of that, perhaps we have extra data that shows the tall cat owners all live on a lush island with the right conditions for both cat breeding and gaining height: so maybe it’s the island that’s causing both cat ownership and average heights to climb.
Alternatively, more cats could be creating a lusher island which in turn is causing people to grow taller – some data on other islands with and without cats would help examine that hypothesis a little more closely.
The more data we have, the quicker we can narrow down the causal relationships.
The only minor wrinkle is with quantum mechanics, where some experiments show correlations that rule out all possible cause and effect relationships… but that’s a story for another day.
Check out the video above and you’ll know just when correlation implies causation, and when it doesn’t.