Talk:Correlation does not imply causation/Archive 2

Archive 1

Archive 2

the relationship between causation and correlation

Generally, causation implies correlation, so the absence of correlation rejects the hypothesis of causation. I think that it's worthwhile mentioning the legitimate interpretation of correlation, rather than just repeating "correlation does not indicate causation" —The preceding unsigned comment was added by 136.142.141.195 (talk) 18:27, 23 February 2007 (UTC).

Welcome to Wiki. If you are going to contribute, please register. Your understanding of the concepts of causation and correlation is poor. You are certainly welcome to comment on talk pages, but please refrain from entering opinion on an article page, unless it is a subject matter which you understand. Further, you should not add anything which for which you can't provide a verifiable source. By the way: Causation does not imply correlation. Correlation suggests the possibility of causation. Absence of correlation might reject an hypothesis of causation, but may also reflect flaws in the model, observation, testing, or interactions of other variables. Your phrase "legitimate interpretation" IMHO is a little scary and evokes images of George Orwell's work such as 1984 and Animal Farm, not to mention political correctness.--Knowsetfree 04:18, 8 May 2007 (UTC)

Um, he didn't say correlation implied causation. He said causation implies correlation, and therefore, a lack of correlation implies a lack of causation. That is correct. If A causes B, then of course A and B will correlate. Conversely, if there is no correlation between A and B, then we can be pretty damned confident that A does not cause B, and vice-versa. As to whether correlation implies causation, that question is a bit trickier. If you define correlation to include apparent correlations that are really just statistical errors, then no, correlation does not imply causation. But if you define correlations to include only those relationships that truly correlate, outside the margin of error, then correlation does indeed imply causation, the only question being whether A causes B, B causes A, or something else causes both A and B. Xrlq (talk) 04:28, 7 January 2009 (UTC)

If A causes B, then of course A and B will correlate.

Nope, that is incorrect. Imagine that X ~ U(-1, 1) and Y = X^2. In this example, correlation between X and Y is 0, yet the causation is total (the value of X determines the value of Y).90.180.172.144 (talk) 16:02, 3 February 2010 (UTC)

The Alice in Wonderland example

Maybe I'm thick, but I can't see how the Alice in Wonderland quote represents an example of the subject of the article at all. What are the two values being correlated, and where is the implied causation? 86.132.59.140 (talk) 20:10, 12 April 2010 (UTC)

image

I noticed that this article lacks images. May I suggest File:PiratesVsTemp English.jpg? bahamut0013^words_deeds 00:57, 23 October 2010 (UTC)

XKCD comic strip

I won't add this to the links since I'm sure it will break some rule I don't know of. But I am very tempted to do so :)

http://xkcd.com/552/

Talgalili (talk) 21:58, 28 December 2010 (UTC)

Example of ideal gas

[...] Given a fixed mass, an increase in temperature will cause an increase in pressure;[...]

Is not strictly correct. The ideal gas law $pV=nRT$ implies that an increase in temperature will cause an increase in pressure provided the volume is held constant, not the mass. The ideal gas law does not make any reference to mass. --86.30.188.213 (talk) 19:01, 21 February 2010 (UTC)

I believe this description while necessary, is still not sufficient and requires further clarification. The ideal gas law $pV=nRT$ requires the quantity, n (not "mass") AND volume to be held constant for a direct correlation to exist between an increase in temperature and corresponding increase in pressure. Otherwise said, "the pressure of a gas of fixed number and fixed volume is directly proportional to the gas's absolute temperature.

This seems to be more an illustration of an indirect cause. The kinetic energy of an ideal gas is a function of (determined by) temperature; meaning, change in temperature "causes" change in kinetic energy. Considering the interior of the container as the system and the walls as the boundary of the system, as energy is (somehow) added to the system through the walls, the temperature of the gas increases causing the kinetic energy of the gas to increase. The increase in kinetic energy causes the gas molecules to collide with the container walls more rapidly / intensely causing the pressure to increase. In re to "not the mass..." if the mass is held constant (or for multicomponent systems if both the mass and composition are held constant) the number (of molecules) will be constant. But I would agree with Hume's observation that in the end we are dependent on observation - and here one cannot directly measure kinetic energy. — Preceding unsigned comment added by Danleywolfe (talk • contribs) 21:21, 2 January 2011 (UTC)

causation requires correlation?

"Correlation is not equal to causation; it is only a requirement for it."

Is it really a requirement? It's conceivable that a causative effect could occur without any noticeable correlation, since some other effect could cancel it out. —Preceding unsigned comment added by 71.167.59.42 (talk) 16:26, 14 December 2009 (UTC)

If it occured and then thing it causes occurs, then that is the correlation; both occured.121.74.4.184 (talk) 03:18, 2 March 2011 (UTC)

The "Coincidence" Example

1. I think that in the given example for "coincidence" the two correlated variables (pirates, CO2) have a common third cause: time. If this is true, we have an example for a negelected confounding variable rather than for coincidental correlation.
2. I am wondering, does the test of statistical significance not rule out coincidental correlation of totally unrelated oberservances by a possibly very high degree of probability? If this is the case, it should not go unmentioned. --80.108.15.242 (talk) 21:41, 27 March 2011 (UTC)

Regarding time as a common third cause: interesting point, but I don't think that time is actually a causative factor here; it is simply another variable that is coincidentally correlated with both CO2 and pirates. Something else that is actually causative underlies CO2, and something else causative underlies pirates. Those underlying causes are themselves correlated with time. But time is just a way of organizing events to see if they are correlated, and does not by itself cause anything.

As for the second point, no, causally unrelated things can evolve similarly in time, and if they are both correlated with time they will be statistically significantly correlated with each other. Duoduoduo (talk) 23:39, 27 March 2011 (UTC)

If the dependent and independent variables are, as you say, "both correlated with time" and, for that reason, "correlated with each other", then time should in my opinion definitely be controlled for.

However, I am increasingly uncertain about my first point. Maybe C02 and pirates are, in fact, not correlated with time: The pirate distribution probably follows something like an inverted U-shaped function and the observed time range just coincides with some corresponding monotonically increasing phase in the CO2 distribution (which itself might follow a choatic pattern).

I think my point about significance is still valid: coincidental correlation is very unlikely, given a high significance level. otherwise, what use would there be in correlational studies? --80.108.15.242 (talk) 10:23, 28 March 2011 (UTC)

That's patently ridiculous. Time isn't a cause; it's a variable. For Christ's sake, you could claim that time is a "common cause" for anything that cooccurs. That's what cooccurence is. Things happening at the same time. Controlling for a time would be like controlling for existence.

--Louiedog (talk) 19:32, 4 May 2011 (UTC)

Louiedog is right that it's tautological that co-occurrence is things happening at the same time. But controlling for time is not "like controlling for existence". In regression analysis, when we want to find out if x causes y but we suspect that both are also caused by separate latent variables that are time-trending, we do a multiple regression of y on both x and time. That way, any effect of x that we pick up can be attributed to the effect that deviations of x, from values associated with the time trend of its latent cause, have on deviations of y from values associated with the time trend of its latent cause. Including time in the regression is called "controlling for time." Duoduoduo (talk) 20:05, 4 May 2011 (UTC)

As much as I enjoyed seeing the FSM pirates as an example, I too feel that it should not be used to demonstrate coincidence because: 1) up until recently there was a third common cause: industrialization of sea-dominant countries; and 2) pirating is currently on the rise, as availability of technology increases further. 94.245.127.13 (talk) 18:23, 4 May 2011 (UTC)

Another example pertinent to current times is the spread of Anthropogenic Global Warming theory (AGW). This real-world example is relevant today. Often science is confused with correlation studies, but science requires more than that-the scientific method.

1. Scientific experiments require a closed system, (or a good approximation of it) to be able to make good assumptions.

2. Logic induction from several experiments that knock out a variable and sets a control (think knockout and control mice).

3. Repeatability.

None of these exist in the AGW theories, because they cannot be tested. First, the biosphere is an open system, full of endogenous variables. A real experiment would by necessity mean to have identical copies of our planet, and systematically remove each of the innumerable variables one by one. (Water vapor, natural CO2 production, plant/algae absorption, sunspot activity, ocean CO2 absorption, fluid dynamic models, etc.) and then account for the effect man-made CO2 has on the biosphere with each of the many variables that affect CO2 and Temperature. In the real world, we only have computer models of our planet, and they are all bad. None can be so complex as to even begin to estimate the effect one of these (let alone all of these) variables have on the biosphere. This is the reality of open systems. There is no ceteris paribus, no knockout variables, no repeatability. There has not been any AGW theories proved using the scientific method, and there never will be. They all depend simply on data observation, and interpretation, and do not meaningfully manipulate the system to measure the changes. In a scientific environment, these theories are as meaningless as coincidences and data correlations are never taken as a measure of causality.

Scientific induction would require more than data correlating temperatures with CO2 concentration and so it is only proper to list it as an example. Although irrelevant to the weight of my argument, I have experience as an Aeronautic Engineer student and a lab assistant to a Crystallographer. I am a political scientist and an economist. 75.89.65.74 (talk) 17:49, 18 May 2011 (UTC)

In the popular environment however, so-called "proven AGW theory" prevails and public policy is made on these causality beliefs. And so Duoduoduo tells me my "opinion disagrees" with others. Sadly, I lost all respect for this self-proclaimed reason man. 75.89.65.74 (talk) 18:15, 18 May 2011 (UTC)

Removing the "coincidence" entirely!

The entire idea of correlation without any casual link of the other types makes no sence... The concept of "coincidence" is that two things happen together by chance. Which means that if you would continue testing, they would not continue to happen together. Which means that there is no correlation, just insufficient statistics... For two things to be correlated it means that you should be able to do "infinitely" many measurements, and consistently get a correlated result.

In the example of the pirates: let's for the sake of the argument assume that there is no "C" that causes both. Then continued measurements for infinitely long time would eventually (even it takes a million years...) show that they are indeed not correlated. Which makes it a flawed example, because there is no correlation.

Compare to a shaman digging a magic stone down next to the field and getting better crops for a few years. Clearly no causation. And for sure no correlation if he would go on sufficiently long with the "measurements". Clearly this example do not belong in this page, as there is neither correlation nor causation. The principle is the same as the pirates.

Seeing the confusion even in the discussion page, should we maybe open another page named "coincidence does not imply correlation"? :)

I'll remove all references to coincidences in a few days unless someone has a better idea. Art.cascade (talk) 16:02, 17 August 2011 (UTC)

"they would not continue to happen together". This is not true. Take any two statistics which are primarily increasing and you'll get repeatedly high correlation again and again. Take, for example, the population of Indonesia and the amount of oxidation on the Statue of Liberty: both are going to be correlating together for a long time,--Louiedog (talk) 19:59, 19 August 2011 (UTC)

yes it's true. I think you don't understand the what I mean with "long time". If you measure one time today, and then every 10000 years, for 1 million measurements (ignoring the fact that malaysia, the statue of liberty and the sun are gone by then), then it is highly unlikely that you will see a correlation, unless there really is a causual link.

What i THINK you mean is that we should measure once a day, or even once a year. Point is that these measurements are not independent. If you are studying something that happens over a timescale of decades, like the change in population in indosia, you cannot do 10 measurements in the same year, and expect to get completely different results. you are just measuring the same thing 10 times. So if I understand you correctly, the correlation you say would persist over a few year is just an artifact of an incorrect statistical analysis. The error here is to assume that the measurements over a few years are independent. All you actually measure is that the two are correlated to themselves over a few years. Yuo get no information on whether they are correlated to each other (or only one sample point).

Most things have a typical timescale that they take to change. some examples:

- position of tennis ball on tennis court. if you measure every 0.1 seconds, the positionts will be havily correalted, but if you measure every 30 seconds, a lot less. If you measure every week, there will be almost none.

- The state of your computers CPU. This change over a timescale of the order of a nanosecond, so if you measure every femtosecond (10^-15), you will se heavy correlation, if you measure every millisecond, you will see no correalation.

I could do more example, burt im sure you get the idea. What I mean with "long time" is a sloppy way of saying that you should measure over much longer time than the typical timescale of thet system in question. If you do not do that, then you can not measure if they actually are correlated or not. I think most of the confusion about this is about people falling into one of the many pitfalls on statistics. :P

I realise that it is not easy to do a proper statistical analysis and error analysis if you are not used to it, and I'm not sure if I'm explaining this thing on the right level. You will proabably feel that I'm going over your head, or that I'm treating you like a baby. :/ Please reply if something is not clear.

Art.cascade (talk) 18:05, 20 August 2011 (UTC)

Before we start, a joke

Never confuse correlation with causation. Since realising this, my life has been so much better.

See also http://xkcd.com/552/

See also http://www.dilbert.com/strips/comic/2011-11-28/ — Preceding unsigned comment added by 75.72.216.32 (talk) 18:28, 29 November 2011 (UTC)

Instance of Informal Fallacy

It would be nice if someone could add something as to what makes this an instance of informal fallacy. In particular, one could address the statement that the logical equivalent of the fallacy is never valid, something which could suggest that this is a formal fallacy. — Preceding unsigned comment added by 78.49.46.35 (talk) 12:43, 1 April 2012 (UTC)

Causal effect and P-value

"causal effect on the disease. This likeliness can be quantified in statistical terms by the P-value". I beleive P-values are only very loosely conncted to this question. --Gaborgulya (talk) 23:58, 25 June 2008 (UTC)

Late to the party here, but agreed. P-values are at best tangential to what's actually being expressed, and even mentioning P-values here is likely to further perpetuate what seems to be a clear and continuous misunderstanding of what P-values actually represent (e.g. P-values can be used as guarantees or worse yet, as probabilities). Although the sentence in question accurately summarizes the measurement in a simple and concise manner, there is no context to express why this value has merit, nor is there any representation of other potential quantified metrics which might meet the same or different conclusions with varying confidence. It would be simply better to express that statistics (with the relevant link) can result in quantification that provides a level of certainty in "such a situation" given the available data. 71.195.177.23 (talk) 06:38, 17 November 2012 (UTC)

Myopia example incorrectly assumes causation

<<a later study at Ohio State University did not find that infants sleeping with the light on caused the development of myopia. It did find a strong link between parental myopia and the development of child myopia, also noting that myopic parents were more likely to leave a light on in their children's bedroom. In this case, the cause of both conditions is parental myopia, and the above-stated conclusion is false.>>

This example makes the same mistake that the whole article is talking about. Correlation is not causation. A 'strong link' does not mean that parental myopia necessarily causes child myopia. In fact, there is every reason to believe that myopia is due much more to near work during childhood than to genetics. In this case, 'parental myopia' only means that the parents read more and use computers more than other people, thus leading their children to adopt the same myopia-causing behaviours. The cause of myopia is not parental myopia but myopia-inducing behaviours in childhood. — Preceding unsigned comment added by WikiWorld88 (talk • contribs) 06:43, 3 August 2013 (UTC)

Suggested Format

The "general pattern" section should be reorganized around the standard threats to casual inference:

1) omitted (unobserved) variables (possibly with a link to the weak page on Mediator variable

2) selection bias

3) simultaneous causation (reverse causation)

4) measurement error

It would be nice to have vivid examples of each type of problem. Cristo00

Indeed. "selection bias" in particular may also interfere with any process relating to the determination of "actual" causation, if such a thing even exists. The best example of this would be the use of inhaled tobacco products correlating to the chances of getting lung cancer. Since correlation is not causation, there assuredly is an infinitude of other possible environmental and genetic contributory factors to investigate, and so the "investigation" phase ("selection bias") assure that in such a case, the inquiry into causation never terminates in anything like a medical or scientific understanding of tobacco-induced lung cancer ever being obtained.

In this way, science can be manipulated. Another way to state this is that science (even math) never actually "proves" anything, or equivalently that all observations, repeatable or not, are anecdotal.

If someone observes that everyone who jumps from a 200 ft high cliff with no parachute dies as a result of the fall, there is always the possibility that such deaths could be blamed on the lack of a trampoline (or an air bag) at the bottom, Newton's law of gravity notwithstanding. Refer also to the excellent Wikipedia article on "ignorance".

Danshawen (talk) 16:09, 27 October 2013 (UTC)danshawen Danshawen (talk) 01:34, 28 October 2013 (UTC)danshawen

Correlation and scientific theory

The basic approach of this article is to suggest that 'causation' is a somehow stronger connection than 'correlation'. So it is claimed it may occur that A and B result from a common cause, but do not cause each other, so a correlation of A and B does not suggest causation.

I doubt the cogency of this approach if A and B always occur together; then calling A the cause of B or vice versa is fine. It may turn out that some event C is connected to A and to B and has a 'deeper' meaning (possibly being a more 'fundamental' mechanism in some sense), but that does not mean C is the real cause. So, for instance, it may be more understandable and more practical to say that a body's fall to the ground is caused by gravity than to attempt an explanation based upon the general theory of relativity as identifying the 'real' cause. Brews ohare (talk) 20:08, 31 March 2014 (UTC)

In the event that A and B occur together only sometimes, then the implication is that there are additional circumstances that have to be identified, and A and B aren't the whole story. The question being raised is how to view causation when we have several ways to 'explain' why A and B always occur together. It would seem that in this case the 'cause' is largely a matter of taste or convenience. Brews ohare (talk) 20:39, 31 March 2014 (UTC)

I think that the point is that you cannot rely on correlation to establish causation, especially in its common means as linear material cause ----Snowded ^TALK 08:35, 2 April 2014 (UTC)

Snowded: I have difficulty understanding your remark. Maybe there is a typo there and you are saying "in its common meaning as 'linear material cause'? Even with that change, I have no understanding of 'linear material cause'. In any event, it is my view that the article has an incomplete discussion of the question of how 'cause' and 'correlation' are related, and ignores the absence of the idea of 'cause' in modern scientific theory. For example, in theories that are time-reversible, the present 'causes' the past as much as the reverse. And the general view is that a 'theory' is simply an economical and practical description connecting various observations, with no claims about 'causes'. In other words, 'cause' is an archaic term (quite possibly demolished by David Hume) and the whole subject of 'cause' vs. 'correlation' is outmoded. Brews ohare (talk) 15:43, 2 April 2014 (UTC)

The point by Brews O'hare is correct. Ultimately causation is just a form for correlation, and this should be pointed out.

Old News

I find it frustrating, annoying, and sad, yet inevitable that this article is named as it is (after a "vote", advertised in the Vogon Town Hall?). Perhaps "that's all I have to say about that" though, as to anyone who does not already understand any explanation would most likely be as pointless as it would be meaningless. What I do want to add though is my belated vote against the US hegemony here onwiki, a US organ despite the hoopla, and for this article to be renamed "Cum hoc ergo propter hoc". The comedy title: "Correlation does not mean causation", could be redirected accordingly. But maybe I am just caught up in "old world" nonsense and should actually assist the process, bravely illustrated here, similarly retitling medical articles such as Pneumonia which of course should be retitled: "Inflammation of the lungs". There are so many fuddy duddy articles to be rebadged. Modus operandi should be given a down to earth title such as "Method of operation" or, if that's still too pompous, then maybe "Way of doing things". The article Ad hoc should become "for this", or maybe "Solution designed for a specific problem or task". Well, there's loads to be done and voting will open soon, on these and thousands of other articles, so check in with Talk pages regularly. Yaay! LookingGlass (talk) 08:21, 19 March 2014 (UTC)

LookingGlass, all the example you gave (Pneumonia, Ad Hoc, Modus operandi, etc.) are used in every day English language and discourse and are the familiar term. The term "Cum hoc ergo propter hoc" is not used in every day language, but "Correlation does not imply causation" is. So I believe keeping the title as it is is the best course of action. I like Salix's comment reply here. Nasa-verve (talk) 20:44, 8 May 2014 (UTC)

A website which contributes to the discussion

Please have a look at this website, which contributes to the discussion with many many examples --Oakwood (talk) 12:38, 20 May 2014 (UTC)

Reference 18, 19, 20 and 21 are blogs

  Bfpage |leave a message  01:05, 29 July 2015 (UTC)

Reference 22

...is an unpublished article proposing a method of determining cause and effect by authors who quote themselves to postulate their new, and as currently unconfirmed method of determining cause and effect. I would characterize this as a primary source quoting a primary source with a serious POV question.

These references are not reliable, notable or secondary sources. The whole last section of this article needs to be removed.

  Bfpage |leave a message  01:19, 29 July 2015 (UTC)

External links modified

Hello fellow Wikipedians,

I have just modified one external link on Correlation does not imply causation. Please take a moment to review my edit. If you have any questions, or need the bot to ignore the links, or the page altogether, please visit this simple FaQ for additional information. I made the following changes:

• Corrected formatting/usage for http://www.economics.soton.ac.uk/staff/aldrich/spurious.pdf

When you have finished reviewing my changes, please set the checked parameter below to true or failed to let others know (documentation at {{Sourcecheck}}).

This message was posted before February 2018. After February 2018, "External links modified" talk page sections are no longer generated or monitored by InternetArchiveBot. No special action is required regarding these talk page notices, other than regular verification using the archive tool instructions below. Editors have permission to delete these "External links modified" talk page sections if they want to de-clutter talk pages, but see the RfC before doing mass systematic removals. This message is updated dynamically through the template {{source check}} (last update: 5 June 2024).

• If you have discovered URLs which were erroneously considered dead by the bot, you can report them with this tool.
• If you found an error with any archives or the URLs themselves, you can fix them with this tool.

Cheers.—^{cyberbot II}_{Talk to my owner:Online} 01:51, 1 April 2016 (UTC)

Missing source for paragraph 2 General pattern

It is necessary to provide a source for the 6 relationships between 2 events. A possible source could be the following book: Causality: Models, Reasoning and Inference by Judea Pearl. As I don't have a copy in my possession, it would be nice if someone cold have a look into this book. I will delete the paragraph in 4 days, if there is no source provided. — Preceding unsigned comment added by Da Vinci Nanjing (talk • contribs) 17:56, 5 February 2017 (UTC)

I believe this section as a whole does not require a source as nobody would deny the fact that correlation can be due to multiple things. However, the question of whether this list is exhaustive could require a source, so I have restored the paragraphed but rephrased it in a way that does not suggest the list is exhaustive. If you think some specific elements of the list would require a source, I am happy to look for one, but I do not think that the general statement requires a source. It seems like an important section of this article, so I think it is important to keep it 7804j (talk) 21:48, 12 February 2017 (UTC)

All material on Wikipedia must be attributable to reliable, published sources, including any analysisis, interpretation, or synthesis such as describing a "general pattern". See § Original research, below. —Sangdeboeuf (talk) 20:59, 29 March 2017 (UTC)

Original research

This article appears to consist of a great deal of original research – most of the general statements about the topic are unsourced, and appear to synthesize the various sourced examples. I'd suggest at a minimum adding the {{original research}} tag until the article has reliable sources to support the interpretation, analysisis, and synthesis given. —Sangdeboeuf (talk) 20:56, 29 March 2017 (UTC)

 Note: I've added {{original research}} to the article. —Sangdeboeuf (talk) 14:52, 19 April 2017 (UTC)

There's a big difference between something being poorly referenced and original research. I've not noticed any original research, although there is much that is unsourced. Given that, and given it's otherwise impossible to remove a tag which is not specific, and given that you have failed to identify any specific problems here, I've removed the general tag; please retag any specific parts of the article you believe may be poorly referenced.GliderMaven (talk) 16:05, 19 April 2017 (UTC)

Move to "Correlation is not causation"

I want to open the discussion on possibly moving this article to "Correlation is not causation". The word "imply" often leads to pointless confusion when referring to this concept, as can be seen in this very talk page. "Is not" is often used and avoids any ambiguity. Does anybody object to this? Karlpoppery (talk) 15:30, 3 May 2017 (UTC)

There must be some historical value to CDNIC, and after all it's more precise in a more formal setting, but I for one support reconciling with CINC. — Preceding unsigned comment added by 2607:FCC8:620C:A000:4422:9DDB:7D2:C445 (talk) 23:39, 23 June 2017 (UTC)

Propositional logic does not belong in article titles. I support this proposal. Michael R Bax (talk) 21:02, 16 November 2017 (UTC)

I am opposed to this move:

• First, the page is about a logical fallacy. It is completely appropriate that the word "imply" is used in its more rigorous sense, in agreement with its use in formal logic.
• Second, the use of this meaning of "imply" is not some rare, weird, jargony usage. To complain that it amounts to "propositional logic in the title" is silly. It is a standard meaning of the word "imply." It is the most natural meaning for the word to have in the context of pointing out a logical fallacy.
• Third, the "does not imply" version is more commonly used than the "is not" version.
• Fourth, some of the core discussion in the article itself would become oddly disconnected from the article title if the change is made. eg: This fallacy is also known as cum hoc ergo propter hoc, Latin for "with this, therefore because of this". The word "therefore" in that version corresponds to the word "imply" in the saying. The fallacy "cum hoc ergo propter hoc" sounds like a subtly different (perhaps vitally different) fallacy than "correlation is not causation."
• Fifth, "correlation is not causation" is imprecise. What would "correlation is causation" even mean? Correlation is something that is directly observed (for instance, in the most formal setting, by statistical analysis of data in an experiment). Causation is something that is inferred or hypothesized or described by theory, about the events that were observed. It only works if the word "is" is suitably interpreted. It is metaphorical use of language, whereas "does not imply" is a precise, logical statement.

— Gpc62 (talk) 00:05, 20 November 2017 (UTC)

I agree with User:Gpc62's points. —Lowellian (reply) 23:29, 13 November 2018 (UTC)

Lice in Middle Ages

In reverse causality example 4 is about lice in health and sick people. There is source (https://blogs.scientificamerican.com/guest-blog/of-lice-and-men-an-itchy-history/) which have been added per citation need a year ago. However, I found no mention of fallacy or anything relating belief. I searched topic and found text (http://meaningring.com/2016/04/08/false-causality-by-rolf-dobelli/) that says story comes from German physics professors Hans-Peter Beck-Bornholdt and Hans-Hermann Dubben. And their book (https://books.google.co.uk/books?id=FnpqAgAAQBAJ&dq=L%C3%A4use+ihren+Wirt,+wurde+er+krank+und+bekam+Fieber&hl=fi&source=gbs_navlinks_s) mention lice:

"Für die Bewohner der Neuen Hebriden war es ganz nat¨rlich, von Läusen besidelt zu sein. Verliessen die Läuse ihren Wirt, wurde er krank und bekam Fieber. Die Menschen waren davon überzeugt, dass man auf einem Fieberkranken immer wieder Läuse aussetzen muss, um das Fieber zu vertreiben. Und in den meisten Fällan gab ihnen der Erfolg Recht. Die Läuse lissen sich wieder auf dem Kranken ansiedeln, und wenig später ging es dem Patienten besser. Allerdings wäre es ihm ohne Läuse wahrscheinlich noch besser gegangen, den hier wurden Ursache und Wirkung verwechselt. Die Läuse verlassen den Kranken, weil er Fiebar hat - sie kriegen ganz einfach heisse Füssen. Und wenn die Hitzewelle vorüber ist, kommen sie gerne wieder (Krämer & Trenkler 1996)."

I don't speak German, but machine translation makes it clear that story is about same but happens in New Hebrides (Neuen Hebriden) instead of Middle Ages. If someone is interested in their source, it can be read here: https://idoc.pub/documents/das-digitale-lexikon-der-populren-irrtmer-8x4e7r83j3l3.Ropertto IV (talk) 11:19, 23 March 2020 (UTC)

Almost incomprehensible section

This paragraph:

"In logic, the technical use of the word "implies" means "is a Sufficient condition for". This is the meaning intended by statisticians when they say causation is not certain. Indeed, p implies q has the technical meaning of the material conditional: if p then q symbolized as p → q. That is "if circumstance p is true, then q follows." In this sense, it is always correct to say "Correlation does not imply causation." In casual use, the word "implies" loosely means suggests rather than requires."

is almost incomprehensible. What is apparently meant to clarify things makes everything much more difficult to understand.

I hope someone knowledgeable in this subject can rewrite this so that readers can benefit from reading it.

Also: In English we do not capitalize adjectives in the middle of a sentence.50.234.60.130 (talk) 19:50, 4 December 2020 (UTC)

Bells Theorem

Bells Theorem does not disprove local causality. This is the mainstream theory right now. What it does is 'disprove' the existence of a Theory exhibiting locality AND counterfactual definiteness.

This would be much more convincing if you had spelled "Bell's" theorem correctly.50.234.60.130 (talk) 19:52, 4 December 2020 (UTC)

Extremely confusing sentence

In general I try to avoid removing citations, in favor of trying to rework problematic cited material. However, here I am removing a citation; this is an explanation of why I did.

The problematic content is the 'sentence' "The widely held (but mistaken) belief that RCTs provide stronger causal evidence than observational studies, the latter continued to consistently show benefits and subsequent analyses and follow-up studies have demonstrated a significant benefit for CHD risk in healthy women initiating oestrogen therapy soon after the onset of menopause.[5]", added in this edit: https://en.wikipedia.org/w/index.php?title=Correlation_does_not_imply_causation&diff=958553497&oldid=958371493 (May 2020 by User:Wikicize).

This sentence is problematic for (at least) two reasons:

• It's grammatically incorrect (the sentence structure is not well-formed)
• It's opinion-based due to the phrase "widely held (but mistaken) belief" which is presented without justification (the previous sentence sources an article that describes a "disparity between findings from observational studies and RCT" which "has created considerable debate among researchers").

Since the linked article (https://www.worldcat.org/title/clinical-pharmacist/oclc/1076797283) is also linked from the article sourced in the previous sentence (https://academic.oup.com/ije/article/33/3/464/716652), it seems reasonable to simply remove the link entirely instead of developing some workaround to maintain it. --Timothy.lucas.jaeger (talk) 01:01, 19 December 2021 (UTC)

After further review, I was mistaken in thinking the reference I was planning to remove was linked from the other sourced article. I will need to rethink my plan of attack here. If someone takes a stab at handling this situation, I would also recommend moving this entire HRT example back to the opening summary from whence it came, as it makes no sense in its current section which is about the meaning of the work 'implies'. This move was done here: https://en.wikipedia.org/w/index.php?title=Correlation_does_not_imply_causation&oldid=958564744 (May 2020 by User:Vsmith). --Timothy.lucas.jaeger (talk) 01:10, 19 December 2021 (UTC)

I decided to simply remove this example entirely. It's hardly a poster child for correlation not implying causation, since there is dispute in the scientific community over whether the observed correlated data or randomly-controlled-trial causal data is actually the correct data to go by (assuming I have any understanding of what any of these cited papers are saying). --Timothy.lucas.jaeger (talk) 01:23, 19 December 2021 (UTC)

Causation does not imply correlation, either.

The article currently states:

Where there is causation, there is correlation, but also a sequence in time from cause to effect, a plausible mechanism, and sometimes common and intermediate causes. While correlation is often used when inferring causation because it is a necessary condition, it is not a sufficient condition

Saying among other things that causation implies correlation. However, I don't think this is true.

Firstly, see https://www.statology.org/does-causation-imply-correlation/ . But secondly, consider this real world example:

I am driving a car on the motorway. The speed limit is 70mph, and I want to get where I'm going as fast as possible, so I'm sticking to the speed limit exactly. The motorway is hilly. When the car is going uphill I press harder on the accelerator to stay at 70, and when the car is going downhill I ease off so that I don't go too fast.

In an analysis of foot-angle vs vehicle speed, you would find no correlation between the two variables, since vehicle speed is constant even though foot-angle varies. However, there is an obvious causal relationship between foot-angle and vehicle speed - the foot-angle controls the accelerator, and the accelerator controls the speed.

This also applies to other control systems, such as room temperature vs AC activity - the AC will aim to maintain a constant temperature, such that you'll get a graph with varying AC activity and a flat room temperature, showing no correlation between AC and temperature, despite obvious causality.

Should I delete the quoted section? FalacerSelene (talk) 20:45, 10 April 2022 (UTC)

No. In your example, there is a confounding variable which cancels a correlation that would otherwise be there. That is a very special case. Also, it is WP:OR, and we do not deleted sourced content because of OR. --Hob Gadling (talk) 05:49, 11 April 2022 (UTC)

Thanks for your explanation. Here we go:

• https://medium.com/gradiant-talks/if-correlation-does-not-imply-causation-then-what-does-8fa462943b84 (the 'does no correlation imply no causation' section)
• https://worthwhile.typepad.com/worthwhile_canadian_initi/2010/12/milton-friedmans-thermostat.html

This wasn't an original thought to me, and I've traced it back to these links. Also, the section that I quoted doesn't have a conflicting citation - it is unsourced. FalacerSelene (talk) 08:42, 11 April 2022 (UTC)