I’ve seen this referenced a lot, especially with psychology, but can someone explain what exactly it means? How does correlation not imply causation? Sometimes, does correlation ever imply causation?
In: 160
A common example used is ice cream sales go up. So does crime at the same time. Ice cream sales drop, so does crime at the same time. Does that mean that ice cream causes crime or crime makes people buy ice cream? Of course not. That’d be a pretty silly thing. Instead there’s another variable going on – the weather. Hot weather increases the desire for ice cream, while it also means more people are out and about, interacting and on shorter fuses.
The other possibility is that they just happen to correlate by pure happenstance (see the previously mentioned Pirates-Global Warming example) and there’s nothing at all joining the two.
Now, if a there is a causation you there WOULD be a correlation; so seeing a correlation may be a signal to a causation, and worth looking into. But showing a causation takes more than just showing correlation.
You could say “Spain implemented this policy to address crime, and crime dropped shortly after”. OK. That’s a potential causation there. But you have to look at more than that. What happened in other countries in that time? Did crime drop in other countries that did not implement the policy? If so, maybe there’s something else going on, like jump in economic growth around the world. It didn’t? Ok. Have other areas implemented similar policies? Did they see similar results?
A statistacly rapport from US said ppl havibg a horse lived longer. The obviois conclusiom is ‘horses is great for your health’. In reality, ppl who can afford horses can normaly afford a good health insurance. Horse+life lengh is a CORROLARION, something here is happening, but not the CAUSEATION. Health care leads to a long life, horses are just expensice so they kind of point out ‘here is someone with money’ who can propably also pay for health insurance without going bankrupt so they get checked out by the doctor in time and fixed before its too late.
ANOTHER: ppl in the mediterian eat a lot of olive oil and drink a lot of wine. They also generaly live a little longer than the avage amarican. It seems there is a CORROLATION, but looking closer its because ppl in eu have free health care, so here it might seem like there is a corrolation but there is not – high wine and oil consumpion actualy decreeses your life expectency, its just health care boosts it even more!
I always carry my wallet when I leave the house. I’ve never been attacked by a tiger outside my house. (Correlation). My wallet protects me from tigers. (Causation)
Correlation = two things happen at the same time, seemingly regularly.
Causation =one thing makes another happen.
So my son gets off the bus every school day at about 3:25.
Everyday at about 3:00 I want to take a nap.
Now, my desire for a nap doesn’t make his bus arrive. His getting out of school (2:50) doesn’t make me tired. They just happen together.
It’s like parallel lines and perpendicular lines. Parallel(Correlation) lines go in the same direction but don’t cross. Perpendicular (Causation )lines cross each other.
Say a friend starts eating a lot of pineapple, and then suddenly he progressively gets promoted. Now he believes that eating pineapple is the cause of his series of promotions.
Correlation does not imply causation. Unless…
Cities with more Churches also have more alcoholics.
There is a big correlation between these two factors.
However they are just correlated. There’s no casual link between these two variables.
The real answer is *larger populations* have more of both.
All Olympic 100m sprint winners wore running shoes. This doesn’t mean however that wearing running shoes will make an Olympic Games 100m winner.
You walk out of your house. A flock of birds flies over head. You crash your car on the way to work.
The birds caused your crash?
Two things that happen may happen at the same time, but be unrelated. Sometimes two things are commonly found together – like there is a bacteria that’s really common in people who have the Flu, so much that it was initially thought to have caused the flu. There was no causal link, though. Just coincidence.
With this, because of this, is a fallacy in logic. Two things can happen over and over and just not be causally linked. One doesn’t have to cause the other, sometimes they just have similar causes.
It’s very common for people to see patterns all the time. We’re really great at patterns. This is why we have so many beliefs about really weird things. Star patterns = who you gonna be based on when in the year you were birthed. Really? Nah. But it’s a pattern.
Correlation is just two data points that seem to match up.
Say you have data on shark attacks, and data on temperature. When looking at the data side by side, you see that as temperature increases, it seems shark attacks do as well. This is called a correlation.
Causation would be saying that increased temperature is what causes sharks to be more agressive based solely on that correl
This may b true, but it may not (i dont actually know, it was just a random example). Correlation does not take into account any other factors. In this case, the shark attacks are likely just because there are more people at the beach when its hot, not that the temp actually affects the sharks in any way. Since the Correlation is there, but it can’t confirm the causation, the idea of Correlation does not equal causation came to be.
However, this does not mean correlations should be dismissed as many people who use that phrase imply, it simply means you need more evidence than just the Correlation to accurately claim causation.
I can actually highly recommend a video by VSauce called “laws and causes” which does not aim to answer this question, but provides very good insight to it regardless.
I think other people have covered a couple angles already, but here’s one slightly different PoV. Imagine we know what caused some effect. Say, we went hunting. At every 30 minutes, we counted how many deer had been killed in the last 30 minutes. And we also measure the temperature of the barrel of the gun. We know that shooting caused the gun to heat up. We also know that shooting caused the deaths of the deer.
So imagine the plot we could make with this data. On the X-axis is the temperature of the barrel and on the Y-axis is the number of deer shot. We would expect that an increase in dead deer would be accompanied by an increase in barrel temperature. This means the two factors are correlated. It may or may not mean they share a common cause. In this example, they do share a common cause. That “CO” in “CO”rrelation in this instance comes from the “CO” in “CO”mmon cause.
But we also know that the hot barrel did not cause the deaths of deer. Nor did dying deer cause the barrel to heat up. They are correlated, but neither one caused the other.
Latest Answers