What's the difference between causation and correlation?

@Satan phil,

Satan;68880 wrote:

Can anyone demonstrate that a cause is more than a mere correlation without engaging in this fallacy?

Are you asking what is the difference between correlation and causation, or are you asking how we know that that that what we believe is causation is not only a correlation? If we can explain why there is a correlation, then we have shown that it is not only a correlation, but also causation. For instance. there is clearly a correlation between the heating of a metal, and the expanding of the metal. And we can explain why heating the metal causes the expansion of the metal. It is because the metal is made up of molecules, and heating causes the molecules to agitate and collide. So, if we can produce a link between the two kinds of correlated events, we can show a causal link.

@kennethamy,

kennethamy;68884 wrote:

If we can explain why there is a correlation, then we have shown that it is not only a correlation, but also causation.

What exactly does it mean for A to "explain why" B happens? What is the difference between a description and an explanation? What's the difference between saying "metal expands after you heat it" rather than saying "metal expands because you heat it"? What's missing in the former that's present in the latter?

@Satan phil,

Satan;68889 wrote:

What is the difference between a description and an explanation? What's the difference between saying "metal expands after you heat it" rather than saying "metal expands because you heat it"? What's missing in the former that's present in the latter?

As I see it, the latter means that the heat makes the metal expand. It implies force or compulsion.

If we have two clocks, and one always strikes the hour 10 seconds after the other, we do not say that one causes the other to strike. This shows that we distinguish between causation and mere correlation.

@ACB,

ACB;68893 wrote:

As I see it, the latter means that the heat makes the metal expand. It implies force or compulsion.

If we have two clocks, and one always strikes the hour 10 seconds after the other, we do not say that one causes the other to strike. This shows that we distinguish between causation and mere correlation.

Ok then I submit that I don't see any such forces. I see heat applied to a piece of metal and then I see a piece of metal expand. I don't see anything more than that. Also, the differences between your clock example and heating a piece of metal are a little bit more than you make out.

When I heat a piece of metal, that piece starts expanding, not some other piece of metal in the next room. Also, pieces of metal don't start expanding absent heat applied directly to them.

On the other hand, your two clocks can chime independantly from each other. I can smash one and the other one still chimes, destroying the correlation completely.

So far, the only difference I've seen is in how strongly and closely correlations are tied. Some are strong, like heating metal and it expanding. Some are weak, like two clocks chiming, or the death of presidents and their election years.

@Satan phil,

The way to distinguish is through statistics.

If after much experimentation (as the smashed clock shows) we still have a 100% synchronicity between the 2 events, with A always preceding B, (in context C) we call it causation.

If there is only a tendency to correlate then we imply that some other event or pirinciple (D) is the underlying cause of both A and B.

In the example of the clock, Newtonian laws are said to cause both the clocks to correlate. When we apply a further newtonian action (a mallet smashing one clock) we can see that newtonian mechanics are not violated, but any idea of the one clock causing the other are violated.

Thus Newtonian laws cause clocks to move.

@Poseidon,

Poseidon;68908 wrote:

If after much experimentation (as the smashed clock shows) we still have a 100% synchronicity between the 2 events, with A always preceding B, (in context C) we call it causation.

Well, I'm pretty sure that smoking cigarettes causes lung cancer but the chances are not 100%.

@Satan phil,

Satan;68880 wrote:

Can anyone demonstrate that a cause is more than a mere correlation without engaging in this fallacy?

Scientific demonstrations don't assert causality as a stone-cold fact. They express results in terms of the probability that an association (aka a correlation) was due to random chance or not.

Let's ask the question of whether beheading causes death, or if it's just a correlation. If I did a cross-sectional study and I found that 100% of people who had been beheaded were dead, that would not prove causality.

If I did a prospective, randomized, controlled trial, and I took 10 people and beheaded half to see what the difference in outcome is -- which would most likely be as you'd expect, then that low number of subjects would not allow me to say that "beheading causes death". What I could say, if the P value were < 0.05, is that there is less than a 5% chance that the association was due to chance alone. If I increase my numbers and compare 1 million beheadings to 1 million controls, then the P value would be so low that one could infer causality since the likelihood of a NON-causal association would be negligible.

@Satan phil,

Satan;68889 wrote:

What exactly does it mean for A to "explain why" B happens? What is the difference between a description and an explanation? What's the difference between saying "metal expands after you heat it" rather than saying "metal expands because you heat it"? What's missing in the former that's present in the latter?

For metal expands after you heat it, all you have to do is observe first heating metal, and then, expanding metal. But when you say, "because" you have a theory like molecular theory to say what happens when you heat metal to the metal which causes its expansion. There are molecules being affected by the heat in a way you can describe. You can, of course, observe two events, cause and effect. But you don't observe, but you postulate the causal connection which explains the effect.

@Satan phil,

I think the syntactic difference is not the explanatory mechanism.

The difference is that with repeated observations you can see that expansion of metal will occur with heating and will not occur without heating, and sufficient observation of this phenomenon (and the sequence of events) proves sufficiently that the heating was causal. This is supported further by the contrary phenomenon of contraction with cooling.

@Satan phil,

Satan;68941 wrote:

Well, I'm pretty sure that smoking cigarettes causes lung cancer but the chances are not 100%.

Well, if you did not inhale the smoke into your lungs, but used the method known locally to me as a 'french inhale' (into the mouth, then straight out the nose), then merely smoking would not cause anything to the lungs.

If you were in the middle of a war-zone, and you did not smoke into the lungs to pacify you, you may lose your cool and get yourself gunned down. In this instance, smoking would prolong your life.

So if you lived in a stressed out city, not smoking may lead you to be more anxious and drive around in your car more as a result, which would mean inhaling fumes, and it may lead to a greater chance of lung cancer.

The point being, that context must always be considered in any association that is not 100%.

@Satan phil,

Smoking does cause over 80% of lung cancers. That is determined by relative risk assessments and odds ratios, i.e. statistical expressions of lung cancer risk as a function of chronic smoking dose.

By the way, Poseidon, don't be so sure that the "french inhale" wouldn't cause lung cancer. Smoking is also a significant independent risk factor for bladder cancer and pancreatic cancer, and smoking is one of the largest of all risk factors for coronary artery disease and stroke -- despite the fact that you're not inhaling directly into any of these body compartments.

@Aedes,

Aedes;68943 wrote:

Scientific demonstrations don't assert causality as a stone-cold fact. They express results in terms of the probability that an association (aka a correlation) was due to random chance or not.

Let's ask the question of whether beheading causes death, or if it's just a correlation. If I did a cross-sectional study and I found that 100% of people who had been beheaded were dead, that would not prove causality.

If I did a prospective, randomized, controlled trial, and I took 10 people and beheaded half to see what the difference in outcome is -- which would most likely be as you'd expect, then that low number of subjects would not allow me to say that "beheading causes death". What I could say, if the P value were < 0.05, is that there is less than a 5% chance that the association was due to chance alone. If I increase my numbers and compare 1 million beheadings to 1 million controls, then the P value would be so low that one could infer causality since the likelihood of a NON-causal association would be negligible.

To say that it's negligible means, for practical reasons, it's useful to ignore it. Yet, causation cannot be proved. It can be only disproved. If you kill the rooster and the sun still comes up then you have disproved that the rooster causes the sun to rise.

@Aedes,

Aedes;68948 wrote:

I think the syntactic difference is not the explanatory mechanism.

The difference is that with repeated observations you can see that expansion of metal will occur with heating and will not occur without heating, and sufficient observation of this phenomenon (and the sequence of events) proves sufficiently that the heating was causal. This is supported further by the contrary phenomenon of contraction with cooling.

What "syntactic difference"? You can observe the correlation of heating with expansion all you like, but unless you are able to say why the two event correlate, you will not be able to say anything about what causes one to be correlated with the other. Just as when we observed how taking an aspirin was followed by the relief of a headache, we still did not know why that sequence of events occurred.

@kennethamy,

kennethamy;68962 wrote:

Just as when we observed how taking an aspirin was followed by the relief of a headache, we still did not know why that sequence of events occurred.

If the correlation between taking aspirin and getting better doesn't explain why you got better then inserting a bunch of medical terminology (which in the end, rests on unexplained chemical and physical properties) doesn't explain it either. Why do apples fall to the ground? The law of gravity. Why is there a law of gravity? Beats me. :perplexed:

@Satan phil,

Satan;68966 wrote:

If the correlation between taking aspirin and getting better doesn't explain why you got better then inserting a bunch of medical terminology (which in the end, rests on unexplained chemical and physical properties) doesn't explain it either. Why do apples fall to the ground? The law of gravity. Why is there a law of gravity? Beats me. :perplexed:

But it isn't medical terminology. It is chemistry. And it explains why aspirin affects headaches. And the chemical properties are not unexplained. They are fairly well understood. Taking the aspirin explain why I got better, that's true. But the question is why taking the aspirin got me better. You are confusing two different question. Why did the my headache get better? Answer, because I took aspirin. And, why did the aspirin cause my headache to get better? Two different questions.

Now, whether or not you know why there is a law of gravity does not mean that it is not true that the law of gravity does not explain why apples fall. You don't have to explain the explanation in order for the explanation to be the explanation. That the glass dropped on the floor explains why the glass broke, whether or not I can explain why the glass dropped.

@kennethamy,

kennethamy;68985 wrote:

But it isn't medical terminology. It is chemistry. And it explains why aspirin affects headaches. And the chemical properties are not unexplained. They are fairly well understood. Taking the aspirin explain why I got better, that's true. But the question is why taking the aspirin got me better.

So you call it explaining something by introducing something else which is unexplained?

Why does aspirin make you feel better. Certain chemical properties interact with your biology. Why do certain chemicals have those properties? Because of covalent and ionic bonding between atoms and their electrons. Why do electrons behave the way they do? We don't know. Why do all electrons have the same charge? We don't know.

Explaining one thing with another that requires it's own explanation is like paying one credit card off with another. Eventually it's going to catch up with you.

@Satan phil,

Satan;69001 wrote:

So you call it explaining something by introducing something else which is unexplained?

I did not say that nothing explains the explanation. I said that an explanation need not be, itself, explained, in order to explain. Why not? Why should I have to explain how the brakes stopped the car in order for it to be true that the car stopped because I applied the brakes? No reason I can see. Would anyone say that when I said the car stopped because I applied the brakes does not explain why the car stopped because I did not, or was not able to explain why applying the brakes stopped the car? Why? And what is going to catch up with me? You seem to think that unless I can explain everything, I have not explained anything. That just seems false.

@kennethamy,

kennethamy;69020 wrote:

You seem to think that unless I can explain everything, I have not explained anything. That just seems false.

So, do you think that you could at least tell me the difference between a true description of X and a true explanation of X? How could I test this difference to see which one I've got?

@Satan phil,

Satan;69030 wrote:

So, do you think that you could at least tell me the difference between a true description of X and a true explanation of X? How could I test this difference to see which one I've got?

A true description would be, "I crossed the street". A true explanation would be, "I crossed the street because I saw my friend there, and I wanted to tell him something". Or a true description would be, "the water in the ice-cube tray turned to ice". The true explanation would be. "The water turned to ice because the temperature in the refrigerator into which I put the tray, was at 0 degrees centigrade".
No problem.

Philosophy Forum

What's the difference between causation and correlation?

Quick Links

My Account

able2know