The whole abortion/breast cancer bruhaha.
I am about as sure as can be that is not true, and causality has been decisively demonstrated. That is why, apart from doing correlational studies on human populations, scientists also did experimental animal studies to demonstrate a true causal link. Possibly more scientific effort went into demonstrating real causality here than into demonstrating causality in any other case. Tobacco companies fought tooth and nail against having warnings printed on cigarette packs and smoking restrictions imposed, on the basis of mere correlations, which is why a lot of scientific effort went into proving real causality.
You do understand, do you, that something can be a cause - even a major cause, even the most important cause, in practice - without being the only significant causal factor, and even without being either a sufficient or necessary cause?
There is really no such thing as THE cause of anything. There are always multiple causal factors for any event.
Certainly smoking is not a sufficient or necessary cause of lung cancer. It does not follow from that that it is not an important cause, and it may be (actually, I am fairly confident it is) true that in most populations with a fairly high incidence of smoking it is the most practically significant cause: the factor that, out of all those you could actually change, if you removed it you would do the most to reduce the incidence of the cancer.
Here’s my favorite example of a spurious correlation:
Write down the number zero on a piece of paper (or a spreadsheet). Flip a coin 100 times. Each time it comes up heads, add one to the previous number; each time it comes up tails, subtract one.
Do this twice, and compute the correlation between the two series. Roughly one sixth of the time the magnitude of the correlations will be greater than 0.7 even though there’s no relationship between them, and no joint cause.
If you attempt to take tripolar’s argument to its logical conclusion, the ONLY thing that causes death is cessation of brain activity. All other supposed causes just correlate with death.
Except, that means tomato consumption and death aren’t correlated in the first place, making it a fairly terrible example of the phrase quoted by the OP, and a good example of how the phrase if often abused.
If I remeber correctly, someone did a correlation analysis (or whatever it is called) of all the statistics that the UN had, and butter production in bangladesh is the best predictor of the US stock market
I found a citefor my assertion. In fact, it turns out it’s not just bangladeshi butter prices. You have to combine butter production in bangladesh, US cheese production, and the sheep population in both countries.
If you do this then you will predict 99% of moves in the S&P. So long as you perform this exercise in the late 80s and early 90s…
A little boy I know thought that only balloons attached to strings would float. He knew that balloons without strings did not float. He said that to keep a balloon from floating away, people should cut off its string.
You can find a bunch of graphical examples on line. From a quick image search –
Import more lemons to decrease traffic fatalities
Well the one I heard years ago which is these days very politically incorrect is that in many cities ice cream consumption has a positive correlation with rape. (IE when ice cream consumption is up so is rape.) From what I remember of that (and it’s been years since I last heard it) basically it amounted to rapes go up in the summer and so does ice cream consumption.
People who eat Count Chocula for breakfast have a lower cancer rate than people who eat oatmeal for breakfast.
The whole point of this thread is to point out the difference between correlation and causation. The scientific evidence clearly shows that smoking does not cause lung cancer. A possible candidate for the cause of the disease is a virus, and smoking likely causes cell damage that allows this virus to infect the lungs. If that is the case, it’s possible that lung cancer vaccine can be developed. If that happened people who smoked would no longer get lung cancer. They would just die from other heart and lung related diseases, and other forms of cancer if those aren’t caused by the same virus. Strong correlations, even extremely strong correlations do not equate to causation.
That is what would be called an illogical conclusion because it has nothing what-so-ever to do with my post.
Do you have a cite for this? I’m not saying you’re incorrect but this really is something I’ve never heard before and I’d be facinated to read more but my Google-fu is really weak.
One famous example is part of the Flying Spaghetti Monster religion, where it correlates the number of pirates with global temperature rise.
Not really much of an example, given that the X-axis looks like it was created by a random number generator.
Even if this is true (and i doubt it is) it remains a perfect example of a causative effect. Causative effects don’t have to be direct to be causative.
In both science and law, what you just described is known as a proximate cause. It’s a cause that is directly linked to, and directly causative of, the immediate cause. It’s a chain of causation.
To give you a more obvious example: Being shot in the heart is a proximate cause of death. The actual cause of death is related to a loss of oxygenation of the brain. It’s perfectly possible to shoot someone who is already on a heart-lung machine and have no effect on their health, so clearly being shot in the heart is not directly causative of death.
But in science, law and common speech, we refer to being shot in the heart as being a cause of death. And even if smoking tobacco causes cancer by increasing the risk of infection by a ubiquitous virus, it is still the *cause *of cancer because there is a direct causative chain. Smoking causes the infection and the infection causes the cancer.
A relationship doesn’t have to be direct to be considered causative. This is a good thing because, as the gunshot example shows, most effects in the real world have an extremely extended chain of causes.
A relationship only becomes *purely *correlative when the two factors are not part of a causal chain. Using an example given earlier: buying ice cream does not in any way lead to being bitten by a snake. It doesn’t cause human behaviour that leads to snakebites. It doesn’t affect the behaviour of snakes. There is simply no chain of causation. It is *purely *correlative because both observations have a common cause: temperature. They don’t cause one another, they both *share *a cause.
That is causality; an indirect cause is still a cause. You can determine whether something is causation or correlation by controlling one of the factors. If you take two groups of people and have half of them stop smoking, they will have less lung cancer, hence there is a cause and effect relationship present. Conversely, in the example of the ice cream sales/homicides correlation, decreasing one will not affect the other.
The fact that theoretically you may be able to separate the cause from the effect by disrupting the method of action doesn’t mean it’s not currently a causal relationship.
Edit: Ninja’ed by Blake. What he said.
If that I so (and I would like to see a cite that this is anything more than a speculative theory) it nevertheless remains the case that smoking causes (and is not merely correlated with) lung cancer. You appear to be quite badly confused about the logic of causation. No event has a single cause. They all have indefinitely many, and a cause of a cause of X is itself a cause of X.
And, in this case, whether or not a virus is involved (and even if the virus is always involved, i.e., it is a necessary cause) it remains the case that, as things are, refraining from smoking will greatly improve your chances of avoiding cancer. It is a cause. The connection is absolutely not mere correlation. (Neither is it what most of the examples in this thread are: a case where two things are correlated because they are both effects of a common cause.)
What you mean, I think, is that smoking is not a sufficient cause of cancer, but that is certainly true whether or not your virus theory is true. It is also not a necessary cause of cancer. (That is, you can get lung cancer even if you never smoke; no doubt even if you are never even exposed to smoke.) Neither of those facts, however, in any way contradict the fact that smoking is a very real, important and significant cause of lung cancer.
The problem with a webcomic level of understanding of the truism that “correlation does not entail causation” is that it too often leads simpler minds to come to the conclusion that “correlation between two variables is never significant.”
Thus people respond to demonstrations of correlations between, say, a policy they do not care for and an outcome that is generally acknowledged to be desirable with an incantation of “correlation does not prove causation.” For instance, “since lowering the legal blood alcohol content level, state X has seen a decline in traffic fatalities,” one argues. “Well, correlation does not prove causation,” our webcomic aficionado retorts.
Now, it is certainly possible for two variables to show correlation while being completely causally unrelated (although this depends on one’s own philosophic understanding as to what it means to say that “C causes E”). In general, using a “common sense” understanding of causality, correlation more often points either to a true cause/effect relationship between two phenomena or that both phenomena are causally related to a third, exogenous common cause (thus ice cream consumption and snakebites are both causally influenced by the advent of summer).
In short, if someone shows you a correlation between two variables, the maxim “correlation is not causation” is the starting point of a retort, but it is deeply insufficient, standing alone, to settle any matter.
Actually I have no argument with that at all. Smoking is one step in a sequence of events that often leads to lung cancer. There is evidence that smoking causes a pre-condition that leads to lung cancer.