50 debunked Science misconceptions

What does make the world turn?
User avatar
Lance Kennedy
Has More Than 9K Posts
Posts: 9869
Joined: Tue Mar 29, 2005 10:20 pm
Custom Title: Super Skeptic
Location: Paradise, New Zealand

Re: 50 debunked Science misconceptions

Postby Lance Kennedy » Mon Sep 18, 2017 7:58 pm

My statement about Xouper being slow had a real point behind it, and I need to explain , since it is obvious this point was missed.

Actually, I should apologise for the implied insult. I was trying to be funny and I guess it missed the mark. Sorry.

The point though, is that Xouper made self contradicting statements. When I accused him of making correlations meaningless, he denied that, and said that the meaning of a correlation was to permit predictions. However, he also said that a large number, probably a majority, of correlations were accidental. That means they happened by chance when random numbers got together. It is very obvious to me, and should be obvious to Xouper, that these two statements put together create an impossibility.

If correlations arise by random chance, they cannot be used for predictions because the next pair of variables in a series will be unrelated. If they are related, then how ?

Which brings me back to the point that solid correlations must have a cause and effect basis. This may be through a third (or fourth etcetera ) variable. Without a cause and effect relationship, there will be no continuing correlation, because that is what accidental or random means. In other words, any correlation based on an accidental juxtaposition of numbers is meaningless, and cannot be used for predictions.

User avatar
xouper
True Skeptic
Posts: 10277
Joined: Wed Mar 30, 2005 5:04 pm
Contact:

Re: 50 debunked Science misconceptions

Postby xouper » Mon Sep 18, 2017 8:53 pm

Lance Kennedy wrote:The point though, is that Xouper made self contradicting statements. When I accused him of making correlations meaningless, he denied that, and said that the meaning of a correlation was to permit predictions. However, he also said that a large number, probably a majority, of correlations were accidental. That means they happened by chance when random numbers got together. It is very obvious to me, and should be obvious to Xouper, that these two statements put together create an impossibility.

If correlations arise by random chance, they cannot be used for predictions because the next pair of variables in a series will be unrelated. If they are related, then how ?

Which brings me back to the point that solid correlations must have a cause and effect basis. This may be through a third (or fourth etcetera ) variable. Without a cause and effect relationship, there will be no continuing correlation, because that is what accidental or random means. In other words, any correlation based on an accidental juxtaposition of numbers is meaningless, and cannot be used for predictions.


How do you propose to demonstrate that correlations that have no causal relation are in fact "random" as you seem to be claiming here? Where is the peer reviewed journal paper that says that?

User avatar
Lance Kennedy
Has More Than 9K Posts
Posts: 9869
Joined: Tue Mar 29, 2005 10:20 pm
Custom Title: Super Skeptic
Location: Paradise, New Zealand

Re: 50 debunked Science misconceptions

Postby Lance Kennedy » Tue Sep 19, 2017 1:51 am

No, Xouper. It was your claim that those correlations were accidental. We all know that some correlations are, indeed, accidental, meaning a random juxtaposition of numbers. All others are from a cause and effect relationship. Some, of course, are causally related through a third variable, such as in Nikki's example of ice creams and murder rate.

But no correlation that is not the result of a cause and effect relationship can be used for making predictions.

User avatar
xouper
True Skeptic
Posts: 10277
Joined: Wed Mar 30, 2005 5:04 pm
Contact:

Re: 50 debunked Science misconceptions

Postby xouper » Tue Sep 19, 2017 2:14 am

Lance Kennedy wrote:No, Xouper. It was your claim that those correlations were accidental.


You have have misinterpreted what I meant.

I did not say that the correlated variables had data with a random distribution.

I said if two variables are selected at random, then some will have a high correlation. That is not the same thing as saying the variables themselves are "unpredictable" or that the data itself is "random".


Lance Kennedy wrote:But no correlation that is not the result of a cause and effect relationship can be used for making predictions.


Where is your journal paper for that claim?

As long as a correlation is reliable, then it can be used to make predictions, whether there is a cause or not.

Example: People who own red cars get into more accidents. Does that mean that the color red is the cause of the accidents? Of course not. And yet insurance companies can predict the accident rate based on the color of the car. They don't even care whether there is a cause or not, the prediction is sufficiently reliable.

User avatar
Lance Kennedy
Has More Than 9K Posts
Posts: 9869
Joined: Tue Mar 29, 2005 10:20 pm
Custom Title: Super Skeptic
Location: Paradise, New Zealand

Re: 50 debunked Science misconceptions

Postby Lance Kennedy » Tue Sep 19, 2017 2:29 am

Red cars and accidents is a causal correlation, via a third variable, which is the fact that more reckless drivers are more likely to buy red cars. More conservative people buy less dramatic colours.

A better example of an accidental correlation is the one I saw on the spurious correlations web site. Over a ten year period, the age of the Miss America finalist correlated quite well with the number of people murdered using steam. That method of murder is very unusual and the number was low. The two variables are so widely different that there is very little chance of any causal relationship. Not even via a third variable. But what stood out to me was that n = 10. Those two statistics would have numbers going almost to the present and right back, probably almost 100 years. So why did the author do the correlation for only ten years ? The answer is obvious. Because it is not a true correlation. As soon as you get outside that ten year period in which by sheer accident a correlation exists, the correlation falls apart. The correlation as presented was the result of cherry picking data, and must be considered dishonest.

The point is that, as long as no causation exists, even via a third variable, you cannot move outside the chain of paired numbers shown on the accidental correlation. Make a prediction, and fall flat on your face. It is no different to accidentally throwing six heads in a row and predicting the seventh will also be heads.

I think that the term 'spurious correlation ' is misleading. Sure, I can do nothing about it, and as long as statisticians accept that term, it will stay. But the word 'spurious ' in common English means fake. And a correlation that works because of the cause and effect through a third variable is not fake. Such correlations are as real as the ones that work by variable A causing variable B. The term 'accidental correlation ' is better, meaning a correlation caused by chance. This kind will not have a large number for n.

User avatar
xouper
True Skeptic
Posts: 10277
Joined: Wed Mar 30, 2005 5:04 pm
Contact:

Re: 50 debunked Science misconceptions

Postby xouper » Tue Sep 19, 2017 2:49 am

Lance Kennedy wrote:Red cars and accidents is a causal correlation, via a third variable, which is the fact that more reckless drivers are more likely to buy red cars. More conservative people buy less dramatic colours.


That's the usual speculation offered, but where is the evidence that is the actual cause, or even that a cause even exists?


Lance Kennedy wrote:The point is that, as long as no causation exists, even via a third variable, you cannot move outside the chain of paired numbers shown on the accidental correlation.

. . . The term 'accidental correlation ' is better, meaning a correlation caused by chance. This kind will not have a large number for n.


Got a journal paper for that claim?

No?

Then you have not met your standard of evidence for your claim.


From this college textbook:

Introduction to Psychology, by Russell A Dewey, 2017 edition
https://books.google.com/books/about/Introduction_to_Psychology.html

Chapter 1, Part 4, Section 2, Correlation and Prediction

. . . Correlations, observed patterns in the data, are the only type of data produced by observational research. Correlations make it possible to use the value of one variable to predict the value of another.

. . . If a correlation is a strong one, predictive power can be great.

. . . Do you need an accurate cause-effect analysis to make predictions?

Correlations can be useful even if we have no theory to explain them. That does not matter if all we want to do is make predictions. All we need is a reliable correlation.

User avatar
xouper
True Skeptic
Posts: 10277
Joined: Wed Mar 30, 2005 5:04 pm
Contact:

Re: 50 debunked Science misconceptions

Postby xouper » Tue Sep 19, 2017 3:27 am

Lance Kennedy wrote:A better example of an accidental correlation is the one I saw on the spurious correlations web site. Over a ten year period, the age of the Miss America finalist correlated quite well with the number of people murdered using steam. That method of murder is very unusual and the number was low. The two variables are so widely different that there is very little chance of any causal relationship. Not even via a third variable. But what stood out to me was that n = 10. Those two statistics would have numbers going almost to the present and right back, probably almost 100 years. So why did the author do the correlation for only ten years ? The answer is obvious. Because it is not a true correlation. As soon as you get outside that ten year period in which by sheer accident a correlation exists, the correlation falls apart. The correlation as presented was the result of cherry picking data, and must be considered dishonest.


Maybe so, maybe not.

You have not shown that it falls apart outside that ten year period. Maybe it does, maybe it doesn't.

Maybe it was cherry-picked, maybe it wasn't. Your conclusion of dishonesty is based on some unsubstantiated assumptions.

In any case, I agree that you will not likely get a reliable correlation for making predictions if you cherry pick the data for the sole purpose of showing a correlation on some small subset. I would not use that particular correlation to predict future murders by steam.

However, if n is large enough, or the sample is shown to be sufficiently representative of the domain (or population of data), then it can be used for predictions, even when there is no cause.

User avatar
Lance Kennedy
Has More Than 9K Posts
Posts: 9869
Joined: Tue Mar 29, 2005 10:20 pm
Custom Title: Super Skeptic
Location: Paradise, New Zealand

Re: 50 debunked Science misconceptions

Postby Lance Kennedy » Tue Sep 19, 2017 3:29 am

No you do not need to have a cause and effects analysis to make predictions. But if a correlation is based on random chance, it will follow the laws of randomness, including the fact that it cannot be used to make predictions. Certainly if n is large enough, the correlation can be used to make predictions. But if n is large, it is not due to random chance.

It is not different to tossing coins. If you tossed a coin 20 times, and it came up heads every time, you know that there is almost certainly a cause and effect reason, such as having a double headed coin. The odds against 20 heads if random are roughly 1 in a million.

User avatar
xouper
True Skeptic
Posts: 10277
Joined: Wed Mar 30, 2005 5:04 pm
Contact:

Re: 50 debunked Science misconceptions

Postby xouper » Tue Sep 19, 2017 3:46 am

Dare I say it seems we have a convergence of understanding here?

User avatar
Lance Kennedy
Has More Than 9K Posts
Posts: 9869
Joined: Tue Mar 29, 2005 10:20 pm
Custom Title: Super Skeptic
Location: Paradise, New Zealand

Re: 50 debunked Science misconceptions

Postby Lance Kennedy » Tue Sep 19, 2017 4:51 am

I dunno.

I have not changed my views, except to allow for changes in definitions. E.g. to define "imply " as leading to a conclusion instead of the normal English word meaning. This is semantics, of course, rather than reality.

My view is still that a strong correlation increases the probability of causation. Noting, of course, that the causation might be via a third variable. Also noting my four conditions for that strong correlation, such as n being a significantly large number. You might recall that I have said from the beginning that n should be 12 or more. Preferably 20 or more.

User avatar
xouper
True Skeptic
Posts: 10277
Joined: Wed Mar 30, 2005 5:04 pm
Contact:

Re: 50 debunked Science misconceptions

Postby xouper » Tue Sep 19, 2017 3:25 pm

Lance Kennedy wrote:No you do not need to have a cause and effects analysis to make predictions. But if a correlation is based on random chance, it will follow the laws of randomness, including the fact that it cannot be used to make predictions. Certainly if n is large enough, the correlation can be used to make predictions. But if n is large, it is not due to random chance.

It is not different to tossing coins. If you tossed a coin 20 times, and it came up heads every time, you know that there is almost certainly a cause and effect reason, such as having a double headed coin. The odds against 20 heads if random are roughly 1 in a million.

xouper wrote:Dare I say it seems we have a convergence of understanding here?

Lance Kennedy wrote:I dunno.


OK, let me clarify. I was referring only to what you said in the quote above.

But it seems you want to change the topic back to one of your previous claims:


Lance Kennedy wrote:My view is still that a strong correlation increases the probability of causation.


OK, then that's a "no".

You have still not met your personal standard of evidence for that claim, despite repeated requests.

Where is the journal paper that says that?

User avatar
Nikki Nyx
Veteran Poster
Posts: 2042
Joined: Wed Jun 07, 2017 12:40 am
Custom Title: cognitively consonant
Location: playing croquet in Wonderland

Re: 50 debunked Science misconceptions

Postby Nikki Nyx » Tue Sep 19, 2017 5:05 pm

xouper wrote:
Nikki Nyx wrote:. . . just having difficulty explaining how I'm thinking about the relationships between the variables, and in what context. I accepted "weak correlation" for salt and pepper, because globally, it is weak. In cultures influenced by French history, there is a strong correlation. But for large parts of the world, the correlation is nonexistent.


That explains the part I was missing. Thanks.
Thank you as well for pointing out that something was missing. I had to dig to find out WHY and HOW my thinking resulted in that conclusion.
What are the facts? Again and again and again-what are the facts? Shun wishful thinking, ignore divine revelation, forget what “the stars foretell,” avoid opinion, care not what the neighbors think, never mind the unguessable “verdict of history”--what are the facts, and to how many decimal places? You pilot always into an unknown future; facts are your single clue. Get the facts!
—Lazarus Long, from Time Enough for Love, by Robert A. Heinlein

User avatar
Nikki Nyx
Veteran Poster
Posts: 2042
Joined: Wed Jun 07, 2017 12:40 am
Custom Title: cognitively consonant
Location: playing croquet in Wonderland

Re: 50 debunked Science misconceptions

Postby Nikki Nyx » Tue Sep 19, 2017 5:32 pm

Lance Kennedy wrote:My statement about Xouper being slow had a real point behind it, and I need to explain , since it is obvious this point was missed.

Actually, I should apologise for the implied insult. I was trying to be funny and I guess it missed the mark. Sorry.
You don't owe me an apology, at least not this time. :mrgreen:

Lance Kennedy wrote:The point though, is that Xouper made self contradicting statements. When I accused him of making correlations meaningless, he denied that, and said that the meaning of a correlation was to permit predictions. However, he also said that a large number, probably a majority, of correlations were accidental. That means they happened by chance when random numbers got together. It is very obvious to me, and should be obvious to Xouper, that these two statements put together create an impossibility.
Could you please quote the two statements you claim contradict each other? Without seeing them in context, I can't really comment, since the above is your interpretation of the statements.

Lance Kennedy wrote:If correlations arise by random chance, they cannot be used for predictions because the next pair of variables in a series will be unrelated. If they are related, then how ?
The next pair of variables in the series may not be unrelated. In fact, it may be a quite long series of variables that establishes the correlation, even though the two events those variables represent are unrelated to each other (except, perhaps, through a third, hidden variable...or not).

Lance Kennedy wrote:Which brings me back to the point that solid correlations must have a cause and effect basis. This may be through a third (or fourth etcetera ) variable. Without a cause and effect relationship, there will be no continuing correlation, because that is what accidental or random means. In other words, any correlation based on an accidental juxtaposition of numbers is meaningless, and cannot be used for predictions.
Even strong correlations can be two events that are unrelated to each other. A and B may have a strong statistical correlation in which C is the cause of both A and B, but A and B remain unrelated to each other except in statistical significance. A is not causing B, and B is not causing A...no cause and effect relationship despite the strong correlation. So I believe you are melding two different concepts together here.

Salt and Pepper have a strong statistical correlation in the United States. The cause of the correlation is the third variable, Louis XIV. But Salt and Pepper remain unrelated to each other, except in statistical significance (their presence together on dining tables in the US). The presence of Salt on the dining table is not the cause of the presence of Pepper and vice versa. No cause and effect relationship exists between Salt and Pepper, even though the correlation is strong.
What are the facts? Again and again and again-what are the facts? Shun wishful thinking, ignore divine revelation, forget what “the stars foretell,” avoid opinion, care not what the neighbors think, never mind the unguessable “verdict of history”--what are the facts, and to how many decimal places? You pilot always into an unknown future; facts are your single clue. Get the facts!
—Lazarus Long, from Time Enough for Love, by Robert A. Heinlein

bobbo_the_Pragmatist
True Skeptic
Posts: 10176
Joined: Wed Sep 09, 2015 9:39 am

Re: 50 debunked Science misconceptions

Postby bobbo_the_Pragmatist » Tue Sep 19, 2017 5:48 pm

Nikki Nyx wrote:No cause and effect relationship exists between Salt and Pepper, even though the correlation is strong.

The casual relationship is concomitant.
Real Name: bobbo the existential pragmatic evangelical anti-theist and Class Warrior.
Asking: What is the most good for the most people?
Sample Issue: Should the Feds provide all babies with free diapers?

User avatar
xouper
True Skeptic
Posts: 10277
Joined: Wed Mar 30, 2005 5:04 pm
Contact:

Re: 50 debunked Science misconceptions

Postby xouper » Tue Sep 19, 2017 6:04 pm

bobbo_the_Pragmatist wrote:
Nikki Nyx wrote:No cause and effect relationship exists between Salt and Pepper, even though the correlation is strong.

The casual relationship is concomitant.


The word "concomitant" is merely a synonym for "strongly correlated". It is not a causal relation.

The salt does not cause the pepper to be there, and the pepper does not cause the salt to be there. There is no causal relation between salt and pepper on your dinner table.

bobbo_the_Pragmatist
True Skeptic
Posts: 10176
Joined: Wed Sep 09, 2015 9:39 am

Re: 50 debunked Science misconceptions

Postby bobbo_the_Pragmatist » Tue Sep 19, 2017 6:10 pm

concomitant. We discussed the various and many forms of "causation" above and that point was not made. I almost looked it up before posting, but lets do it now:

1. Following or accompanying as a consequence
2. An event or situation that happens at the same time as or in connection with another

Does not "say" its a form of causation, it just reeks of it, plus.... thats what I was taught in high school...... so: your authority is what?
Real Name: bobbo the existential pragmatic evangelical anti-theist and Class Warrior.
Asking: What is the most good for the most people?
Sample Issue: Should the Feds provide all babies with free diapers?

User avatar
xouper
True Skeptic
Posts: 10277
Joined: Wed Mar 30, 2005 5:04 pm
Contact:

Re: 50 debunked Science misconceptions

Postby xouper » Tue Sep 19, 2017 7:11 pm

bobbo_the_Pragmatist wrote:concomitant. We discussed the various and many forms of "causation" above and that point was not made. I almost looked it up before posting, but lets do it now:

1. Following or accompanying as a consequence
2. An event or situation that happens at the same time as or in connection with another

Does not "say" its a form of causation, it just reeks of it, plus.... thats what I was taught in high school...... so: your authority is what?


Before I made my previous post, I had thought it prudent to double check my understanding by looking at three different online dictionaries. None of them had definition #1 above, only definition #2, which is a synonym for "correlation". I quit looking after that (my confirmation bias having been satisfied), but apparently that was a mistake, for it seems now that if I had kept looking, I might have found the dictionary you copied from, so I'll concede that point in your favor.

In any case, your definition #1 does not apply to salt and pepper. The salt does not cause the pepper to accompany it, and vice versa. There is no causal relation between the salt and pepper on your dinner table.
Last edited by xouper on Tue Sep 19, 2017 7:44 pm, edited 1 time in total.

bobbo_the_Pragmatist
True Skeptic
Posts: 10176
Joined: Wed Sep 09, 2015 9:39 am

Re: 50 debunked Science misconceptions

Postby bobbo_the_Pragmatist » Tue Sep 19, 2017 7:36 pm

I use the free and very useful https://wordweb.info/. Hotkey for quick look ups. Not indepth, could be much better, but free and useful.

#2 is NOT a synonym for correlation as per the contextual meaning of "connection"

I'm going to bet myself that a quick google of (types of causation) will include concomitant...............................well, THAT was certainly unsatisfying. Pretty good discussion at https://en.wikipedia.org/wiki/Causality ... nditionals that does mention the word but not as desired. Hmmmm....could Mr Snyder in Basic Chem have gotten it wrong?????

https://en.oxforddictionaries.com/defin ... Definition of concomitant - naturally accompanying or associated. //////That means its not a "mere" correlation.......right? I still like my own first definition given above: when A causes B&C then B&C are concomitantly caused by A.

None of the 30 definitions reviewed speak against such a contextualized definition.
Real Name: bobbo the existential pragmatic evangelical anti-theist and Class Warrior.
Asking: What is the most good for the most people?
Sample Issue: Should the Feds provide all babies with free diapers?

User avatar
xouper
True Skeptic
Posts: 10277
Joined: Wed Mar 30, 2005 5:04 pm
Contact:

Re: 50 debunked Science misconceptions

Postby xouper » Tue Sep 19, 2017 7:54 pm

bobbo_the_Pragmatist wrote:I still like my own first definition given above: when A causes B&C then B&C are concomitantly caused by A.


OK, I will accept that interpretation.

That does not change the fact that there is no causal relation between salt and pepper.

The phrase "causal relation between B and C" (as used in this context by the scientific community), does not include being caused by a third variable A.

Since we are having a conversation about a scientific matter, then it is reasonable to insist that the jargon be used as the scientific community defines it, not how Lance defines it.

bobbo_the_Pragmatist
True Skeptic
Posts: 10176
Joined: Wed Sep 09, 2015 9:39 am

Re: 50 debunked Science misconceptions

Postby bobbo_the_Pragmatist » Tue Sep 19, 2017 7:59 pm

Yes.............thats because the casual relationship is BETWEEN B and A, and C and A. NOT BETWEEN B&C EVEN THOUGH there is a strong correlation between the two. If that relationship is not (concomitantly) casual, how would you better describe it?
Real Name: bobbo the existential pragmatic evangelical anti-theist and Class Warrior.
Asking: What is the most good for the most people?
Sample Issue: Should the Feds provide all babies with free diapers?

User avatar
xouper
True Skeptic
Posts: 10277
Joined: Wed Mar 30, 2005 5:04 pm
Contact:

Re: 50 debunked Science misconceptions

Postby xouper » Tue Sep 19, 2017 8:03 pm

bobbo_the_Pragmatist wrote:Yes.............thats because the casual relationship is BETWEEN B and A, and C and A. NOT BETWEEN B&C EVEN THOUGH there is a strong correlation between the two. If that relationship is not (concomitantly) casual, how would you better describe it?


I accept your description.

You are merely saying that B and C have a third cause, A. I have no problem with that.

User avatar
Lance Kennedy
Has More Than 9K Posts
Posts: 9869
Joined: Tue Mar 29, 2005 10:20 pm
Custom Title: Super Skeptic
Location: Paradise, New Zealand

Re: 50 debunked Science misconceptions

Postby Lance Kennedy » Tue Sep 19, 2017 8:20 pm

This whole discussion is about semantics.

Initially I disputed the slogan that correlation does not imply causation, since by the normal meaning of the word imply (suggest) it often does suggest causation . However, if we use the non normal English definition of imply to mean 'leads to a conclusion ', then the slogan is correct. Xouper is suggesting that the salt and pepper on tables are not correlated. Wrong. He is using the wrong words. They are not directly correlated, but are most definitely correlated via an indirect method. Without specifying direct, he is misleading.

There are four kinds of correlation, between two variables, A and B.
1. A causes B.
2. B causes A.
3. Third variable C causes both.
4. The correlation is accidental. Ie due to random chance.

The first three are due to cause and effect.

Correlations that are due to random chance cannot be used for prediction. If there are ten pairs of variables that correlate due to random chance, the probabality that the eleventh pair will also correlate is purely down to random chance also. Thus you cannot use the first ten to predict the eleventh. Only correlations that are based on cause and effect can be used for predictions.

bobbo_the_Pragmatist
True Skeptic
Posts: 10176
Joined: Wed Sep 09, 2015 9:39 am

Re: 50 debunked Science misconceptions

Postby bobbo_the_Pragmatist » Tue Sep 19, 2017 8:58 pm

Lance: Agreed.

I also note the use of quibble. Its worthwile when it brings additional, deeper, or different meaning to a subject. Just a pile of {!#%@} when it obfuscates.
Real Name: bobbo the existential pragmatic evangelical anti-theist and Class Warrior.
Asking: What is the most good for the most people?
Sample Issue: Should the Feds provide all babies with free diapers?

User avatar
xouper
True Skeptic
Posts: 10277
Joined: Wed Mar 30, 2005 5:04 pm
Contact:

Re: 50 debunked Science misconceptions

Postby xouper » Tue Sep 19, 2017 9:17 pm

Lance Kennedy wrote:This whole discussion is about semantics.


It seems that way, mostly.


Lance Kennedy wrote:Initially I disputed the slogan that correlation does not imply causation, since by the normal meaning of the word imply (suggest) it often does suggest causation .


The word "suggest" is still in dispute here, otherwise, I agree with your observation.


Lance Kennedy wrote:However, if we use the non normal English definition of imply to mean 'leads to a conclusion ', then the slogan is correct.


If by "non normal English definition" you mean "the scientific and mathematical definition", then yes, what you said is correct.


Lance Kennedy wrote: Xouper is suggesting that the salt and pepper on tables are not correlated. Wrong. He is using the wrong words. They are not directly correlated, but are most definitely correlated via an indirect method.


I suspect you did not mean to use the word "correlate" as you did. I suspect you meant "causally related".

I have never said that salt and pepper were not correlated. I have very clearly said they ARE correlated.

I assume then, that you accidentally typed "correlation" and that you probably had something else in mind.

Typos happen. Been there done that.


Lance Kennedy wrote: Without specifying direct, he is misleading.


In the scientific definition of the term "causal relation", the word "direct" is always assumed. It is part of the definition. I assumed you knew that.

I was very clear in stating that a "causal relation between A and B" is defined by the scientific community to mean only that A caused B, or that B caused A.

It was not my intention to mislead anyone. My point here is that in a discussion of scientific ideas, it is proper to insist on using the jargon correctly. Otherwise, you risk equivocation errors.


Lance Kennedy wrote:There are four kinds of correlation, between two variables, A and B.
1. A causes B.
2. B causes A.
3. Third variable C causes both.
4. The correlation is accidental. Ie due to random chance.

The first three are due to cause and effect.


The scientific community defines "causal relation" as being only #1 or #2.

You do not have the authority to redefine what the scientific community means by the term "causal relation".


Lance Kennedy wrote:Correlations that are due to random chance cannot be used for prediction. If there are ten pairs of variables that correlate due to random chance, the probabality that the eleventh pair will also correlate is purely down to random chance also. Thus you cannot use the first ten to predict the eleventh. Only correlations that are based on cause and effect can be used for predictions.


You have conflated two kinds of "random" in your comments. I have explained this before.

There are two different ways #4 can be "random".

One way is that A and B are chosen at random (or by accident), but A and B are not comprised of random numbers.

The second way, A or B or both could be sets of numbers that seem to be random.

Your comments about prediction apply only to the case where A or B are a set of random numbers.

Thirdly, contrary to what you seem to be saying, it is not always the case in #4 that A and B are sets of random numbers. Some examples of #4 have data that are not random.

Fourthly, in a previous post, I cited a college textbook that clearly states when a correlation can be useful for prediction even when there is no causal relation. Whereas you have cited nothing that meets your own standard of evidence.

In a previous post, I have already agreed with your point about correlations which are cherry picked, or where n is too small, or the data are random numbers.

But when n is sufficiently large, or the sample can be shown to be representative of the whole, then such a correlation can be used for predictions, even when there is no causal relation.

User avatar
xouper
True Skeptic
Posts: 10277
Joined: Wed Mar 30, 2005 5:04 pm
Contact:

Re: 50 debunked Science misconceptions

Postby xouper » Tue Sep 19, 2017 9:22 pm

bobbo_the_Pragmatist wrote:Lance: Agreed.

I also note the use of quibble. Its worthwile when it brings additional, deeper, or different meaning to a subject. Just a pile of {!#%@} when it obfuscates.


My purpose here is to prevent equivocation errors. It is not my intention to obfuscate.

In a discussion of scientific issues, it is entirely proper to insist on using the jargon correctly. Anything less leads to semantic ambiguity, and possibly obfuscation.

If you are truly interested in avoiding obfuscation, then use the jargon properly.

bobbo_the_Pragmatist
True Skeptic
Posts: 10176
Joined: Wed Sep 09, 2015 9:39 am

Re: 50 debunked Science misconceptions

Postby bobbo_the_Pragmatist » Tue Sep 19, 2017 9:51 pm

Define equivocation or equivocation errors if it changes meaning when being compounded.

"I do not think that word means what he thinks...."
Real Name: bobbo the existential pragmatic evangelical anti-theist and Class Warrior.
Asking: What is the most good for the most people?
Sample Issue: Should the Feds provide all babies with free diapers?

User avatar
xouper
True Skeptic
Posts: 10277
Joined: Wed Mar 30, 2005 5:04 pm
Contact:

Re: 50 debunked Science misconceptions

Postby xouper » Tue Sep 19, 2017 10:02 pm

:roll:

bobbo_the_Pragmatist
True Skeptic
Posts: 10176
Joined: Wed Sep 09, 2015 9:39 am

Re: 50 debunked Science misconceptions

Postby bobbo_the_Pragmatist » Tue Sep 19, 2017 10:13 pm

Using emojis is equivocation by any definition.
Real Name: bobbo the existential pragmatic evangelical anti-theist and Class Warrior.
Asking: What is the most good for the most people?
Sample Issue: Should the Feds provide all babies with free diapers?

User avatar
Lance Kennedy
Has More Than 9K Posts
Posts: 9869
Joined: Tue Mar 29, 2005 10:20 pm
Custom Title: Super Skeptic
Location: Paradise, New Zealand

Re: 50 debunked Science misconceptions

Postby Lance Kennedy » Tue Sep 19, 2017 10:15 pm

Xouper

Again semantics rears its ugly head.


You have referred to correlations that are due to a third factor as not being causally related. Yet it is a cause and effect relationship. 'A' may not cause 'B' or the reverse. But if both are caused by factor 'C', then the relationship is cause and effect, even if indirect. Such a correlation can be used for prediction, since it is cause and effect based.

But if a correlation is due to accident, then no matter how big n may be, it cannot be used for prediction. Of my four types, only 1, 2, and 3 can be used for prediction. Mind you, if a correlation is accidental, n will normally be small, due to the fact that the probability of more variables being correlated drops with more being considered. If you throw a coin ten times, the probability of ten heads is 1 in 1000 roughly. If you throw 20 times, the probability of 20 heads is 1 in a million. The number of coin tosses can be considered to be similar to the number n in a correlation. If the result is due to random chance, then the higher the number of coin tosses, or n, the lower the chance of continued consistent results.

As I said, semantics is ugly, when it obscures the underlying reality. I find the term "spurious correlation " to be such an obscuring piece of semantics, since the word spurious means fake. A correlation due to a third factor is not fake. It is a real correlation, and can be used to make predictions. It is kind of sad that some idiot mathematicians chose such a misleading term.

User avatar
xouper
True Skeptic
Posts: 10277
Joined: Wed Mar 30, 2005 5:04 pm
Contact:

Re: 50 debunked Science misconceptions

Postby xouper » Tue Sep 19, 2017 10:18 pm

bobbo_the_Pragmatist wrote:Define equivocation or equivocation errors if it changes meaning when being compounded.

"I do not think that word means what he thinks...."


https://en.wikipedia.org/wiki/Equivocation

User avatar
xouper
True Skeptic
Posts: 10277
Joined: Wed Mar 30, 2005 5:04 pm
Contact:

Re: 50 debunked Science misconceptions

Postby xouper » Tue Sep 19, 2017 10:22 pm

Lance Kennedy wrote:Xouper

Again semantics rears its ugly head.


Apparently so.

In a discussion about scientific matters, which definitions shall we use? Do we use the jargon the way the scientific community defines it, or your way?

Answer: not your way.

User avatar
xouper
True Skeptic
Posts: 10277
Joined: Wed Mar 30, 2005 5:04 pm
Contact:

Re: 50 debunked Science misconceptions

Postby xouper » Tue Sep 19, 2017 10:23 pm

Lance Kennedy wrote:But if a correlation is due to accident, then no matter how big n may be, it cannot be used for prediction. Of my four types, only 1, 2, and 3 can be used for prediction.


Where is your journal paper that says that?

bobbo_the_Pragmatist
True Skeptic
Posts: 10176
Joined: Wed Sep 09, 2015 9:39 am

Re: 50 debunked Science misconceptions

Postby bobbo_the_Pragmatist » Tue Sep 19, 2017 10:24 pm

Oh..........the equivocation fallacy........interesting how the common English meaning of the word changes as it is combined with fallacy.

My view: you mostly engage in the equivocation fallacy. Concept has two meanings. You give meaning A as if it was the only definition that applies but do begrudgingly admit to definition B if one is offered....but then your position reverts to using only definition A. So...you agree and disagree at the same time. A robust misuse of the fallacy.
Real Name: bobbo the existential pragmatic evangelical anti-theist and Class Warrior.
Asking: What is the most good for the most people?
Sample Issue: Should the Feds provide all babies with free diapers?

User avatar
xouper
True Skeptic
Posts: 10277
Joined: Wed Mar 30, 2005 5:04 pm
Contact:

Re: 50 debunked Science misconceptions

Postby xouper » Tue Sep 19, 2017 10:36 pm

bobbo_the_Pragmatist wrote:My view: you mostly engage in the equivocation fallacy.


I don't give a rat's ass what your view is.

bobbo_the_Pragmatist
True Skeptic
Posts: 10176
Joined: Wed Sep 09, 2015 9:39 am

Re: 50 debunked Science misconceptions

Postby bobbo_the_Pragmatist » Tue Sep 19, 2017 10:40 pm

Yes, yes........was that an attack? Can't really tell with you.

Heh, heh. What are you saving all those rat's's's asses for?
Real Name: bobbo the existential pragmatic evangelical anti-theist and Class Warrior.
Asking: What is the most good for the most people?
Sample Issue: Should the Feds provide all babies with free diapers?

User avatar
Lance Kennedy
Has More Than 9K Posts
Posts: 9869
Joined: Tue Mar 29, 2005 10:20 pm
Custom Title: Super Skeptic
Location: Paradise, New Zealand

Re: 50 debunked Science misconceptions

Postby Lance Kennedy » Tue Sep 19, 2017 11:41 pm

https://www.jstor.org/stable/2985494?se ... b_contents

Quoting from the paper above. Note that R refers to the correlation coefficient.

"R has come to be accepted, at least in economic and marketing studies, as a major criterion of the adequacy of a hypothesis. "

In other words, a good correlation is meaningful in terms of determining causation.

What is a good correlation ?

The equation for this (t is the reliability of a correlation)

T = R times square root of ((n-2) divided by (1 - R))

This equation makes it clear that there are only two factors determining how reliable a correlation is. The coefficient and the number n.
You may recall that these two are two of the conditions I laid down for a correlation with a high probability of causation. R should be more than 0.5 and n should be more than 20. T is also the value of the correlation in determining predictions. If t is low, say 2 or less, then the correlation is pretty much worthless.

User avatar
xouper
True Skeptic
Posts: 10277
Joined: Wed Mar 30, 2005 5:04 pm
Contact:

Re: 50 debunked Science misconceptions

Postby xouper » Wed Sep 20, 2017 12:20 am

Lance Kennedy wrote:https://www.jstor.org/stable/2985494?seq=1#page_scan_tab_contents

Quoting from the paper above. Note that R refers to the correlation coefficient.

"R has come to be accepted, at least in economic and marketing studies, as a major criterion of the adequacy of a hypothesis. "

In other words, a good correlation is meaningful in terms of determining causation.


That's not a valid interpretation of what he wrote.

What that means is that if you already have evidence for a causal hypothesis, then the correlation has been used (by some) as a measure of its adequacy, but not necessarily as evidence for the existence of the cause. The existence of the cause must be established before applying R as a measure of the adequacy of the hypothesis.

He also says it is not necessarily valid to use R that way.

He is also not talking about science. Those in economics and marketing do a lot of things that are not scientific.

Try again.

User avatar
xouper
True Skeptic
Posts: 10277
Joined: Wed Mar 30, 2005 5:04 pm
Contact:

Re: 50 debunked Science misconceptions

Postby xouper » Wed Sep 20, 2017 12:26 am

Again:

Lance Kennedy wrote:But if a correlation is due to accident, then no matter how big n may be, it cannot be used for prediction. Of my four types, only 1, 2, and 3 can be used for prediction.


Where is your journal paper for that claim?

Ferber's paper, which you cited in a previous post, does not not even address that claim, let alone support it. Try again.

User avatar
Lance Kennedy
Has More Than 9K Posts
Posts: 9869
Joined: Tue Mar 29, 2005 10:20 pm
Custom Title: Super Skeptic
Location: Paradise, New Zealand

Re: 50 debunked Science misconceptions

Postby Lance Kennedy » Wed Sep 20, 2017 12:58 am

Xouper

Anything based on random chance cannot be used for prediction. As I said before, tossing ten heads in a row does not change the probability of the eleventh toss. The same principle applies to correlation. If ten pairs of variables are correlated by random chance, the eleventh does not have to be correlated. It is a simple principle of probability.

Only if cause and effect is involved can something be predictable. Random chance by its basic nature is not predictable. That is what random means.

User avatar
xouper
True Skeptic
Posts: 10277
Joined: Wed Mar 30, 2005 5:04 pm
Contact:

Re: 50 debunked Science misconceptions

Postby xouper » Wed Sep 20, 2017 1:11 am

Lance Kennedy wrote:Xouper

Anything based on random chance cannot be used for prediction.


I already agreed with that.

What you seem to be claiming is that any correlation in category #4 must be based on random chance.

Where is your journal paper for that claim?


Return to “Science, Technology, and Mathematics”

Who is online

Users browsing this forum: No registered users and 1 guest