Misplaced Pages

Revision as of 07:26, 21 January 2006

a struggle for understanding

":The only time a sample size of 1 is statistically valid is when you're disproving a conclusion that claims to be correct for 100% of the population in question (which really only happens in logical arguments, not statistical arguments). So, if you're trying to make the argument that cigarettes don't cause cancer, the argument is flawed because the statistical evidence you're arguing from is too weak to support the conclusion. However, the exact same anecdotal evidence could be used to support a very strong argument that cigarettes don't always cause cancer, so to say that an argument is fallacious just because the evidence it's based on is anecdotal is clearly wrong. --Flatline 17:05, 3 January 2006 (UTC)"

All single statistic points are valid if gathered factually. The conclusion gained from the statistic is what is invalid

"gathered factually" and "statistically valid" are two very different things. For a single data point to be statistically valid, you have to be able to show what population the data point came from (which can be very difficult to do after the sample has been taken), demonstrate that the sample was randomly chosen from that population to prevent bias (impossible to show in the case of an anecdote), and, since your sample size is so small, you need to be able to make some sort of argument that the population in question is homogeneous (impossible to make without taking more samples). --Flatline 16:37, 14 January 2006 (UTC)

>Statistically valid" used that way, is not used on the evidence, it is used on the conclusion. The conclusion isn't statistically valid because the statistics are incomplete, not because the evidence is invalid. If a statistic is invalid it is incorrect. All statistics are valid unless incorrect. The statistics gathered may be valid but insufficient to produce any valid statistically conclusion. <

--Eric Norby 11:11(PST), 2006, January 17 (AD)

And as you say it can be valid if you are using the single datum to point out that not 100 percent of the points are inside some conclusion.

In that case, the datum is being used as a counter example, not as statistical evidence. It is statistically invalid. I was mistaken when I called it statistically valid in a previous comment. --Flatline 17:45, 15 January 2006 (UTC)

>The datum used in a counter example will only be invalid if it is a "flyer", a point that may have come from measurement or experimental error. That is determined at the time of the test/sample. If the statistic is undeniably correct, it is not a flyer, it is a valid statistic. Counter examples are used to invalidate a statistically incorrect conclusion< --Eric Norby 11:11(PST), 2006, January 17 (AD)

It is not the evidence that is invalid. It is the conclusion.

A person falling from a height of 2000 feet up didn't die. Does that mean that falling from great height doesn't lead to great death? No but it is valid that falls from great height don't kill everyone.

It is anecdotal to say that falls from great height don't kill.

I've never heard "anecdotal" used this way before. You have some anecdotal evidence that falls from great heights don't kill, but you could just as easily have anecdotal evidence that says falls from great heights do kill, so to apply the word "anecdotal" to the conclusion doesn't tell me anything about the correctness of the conclusion. I did a google search on "anecdotal conclusion" and found lots of examples of people using the phrase to mean "conclusion based on anecdotal evidence". Would I be correct if I said that you would claim that it means "conclusion that is verifiably false but supported by (carefully chosen) correct evidence that misrepresents the true state of things"? --Flatline 17:41, 15 January 2006 (UTC)

>My example was a rewrite of the cancer one. I'm going to try another rewrite of the current definition from what someone else has put forth. Anecdotal evidence should be short data that supports a point. Like anecdote means short story to support a point. It has come to mean short evidence that appears to support but doesn't a conclusion. That conclusion may be false. < --Eric Norby 11:11(PST), 2006, January 17 (AD)

Is it statistically valid that I dropped a hammer and it fell. Yes.

It is a fact that you dropped a hammer and it fell. It is meaningless to ask if it's statistically valid without first identifying "for what purpose?". There is a difference between a fact and a statistic; they are not the same thing. Similarly, there is a difference between scientific data and mere collection of facts (hence the cliche: "The plural of anecdote is not data"). --Flatline 17:41, 15 January 2006 (UTC)

>Exactly. Statistical validity is a term used on a conclusion, not on the statistics. There is no difference between statistics and facts. If taken using proper scientific tools both will be scientific facts. Using facts in statistical mathematics is using facts statistically. Statistics is a collection of scientific facts.

I suppose you have a point that there are "Lies, dam lies and statistics." ;)

What do you consider to be the definition of a statistic?< --Eric Norby 11:11(PST), 2006, January 17 (AD)

Is it wrong to say, "Therefore gravity must still exist." No. Does the single datum give conclusive evidence to that statement? Yes and no. It does support it, and isn't conclusive.

Being a single datum doesn't lead to statistically erroneous data. Using a single point in an attempt to support an incorrect conclusion, is using the data anecdotally, and a weak statistical conclusion. The datum is still correct and valid.

The statistical evidence isn't too weak to support the conclusion. The conclusion is weak because the evidence given doesn't fully support it. The evidence is just as correct as it always was. There is nothing wrong with the evidence.

Anecdote is a short story given to support a larger paper. Anecdotal evidence is supposed to be short evidence to support an anecdote. It has turned into, short evidence that doesn't support an anecdote.

"Charlie's grandmother lived to 95", is anecdotal if you use it to support the point that smoking doesn't cause cancer, however, it isn't anecdotal if used to support the point that not all smokers die of lung cancer.

For you information, that evidence doesn't support "that cigarettes don't always cause cancer", because that person may have had cancer but had not yet grown to significance, or if she had lived long enough she would have gotten it.

" :I agree that it is incorrect to say that anecdotal evidence is pseudoscientific. It would be more correct to say that anecdotal evidence is unscientific. However, generally speaking, it is correct to say that it is statistically unverified since if the anecdote were statistically verified, the argument would be based instead on the far stronger statistical evidence and not the anecdotal evidence. --Flatline 17:05, 3 January 2006 (UTC) "

How can a statistic be "statistically unverified". Is it not a statistic? Is it not verifiable? The evidence given is absolutely correct. The error is assuming it supports something that it doesn't.

" :You're correct: the conclusion is too strong for the evidence given. But this does not support your argument that anecdotal evidence necessarily supports fallacious conclusions. --Flatline 17:05, 3 January 2006 (UTC) "

But it does support my conclusion that the "support" is fallacious, therefore the conclusion is unwarranted.

It is a fallacy of support. The evidence is correct, the conclusion often is not, and the support is fraudulent. It is a scientific fallacy of fraudulent support, assuming logical support where none is apparent.

" :It is a false dichotomy to claim that data can't be both anecdotal and false. --Flatline 17:05, 3 January 2006 (UTC) " A false dichotomy is where two choices are given when there are three or more.

You should either call a fallacy by its most correct name or not. I don't see a valid reason for calling a fallacy by more than one name.

If you use false data for supporting an anecdote, it is much more correct to point out the falsehood of the data.

"Charlie's grandmother died at age 95....etc.", if that data were incorrect, it wouldn't support the anecdote. The anecdote would be left null and void. It would be far wiser to point out that Charlie's grandmother was diagnosed with cancer a week before she was hit by a car, and her health was poor a result of heart problems and emphysema.....and her lungs were found full of charcoal, ash, and tar, than it would to say, "that is anecdotal". False data doesn't make it to the anecdotal fallacy stage.

Sure it can be used incorrectly that way, but the best and most correct defense is derived from false data fallacy rather than the anecdotal fallacy.

The anecdotal fact must be indisputable by both parties, or the evidence will be under attack, the support lost from the beginning. Good data that appears to support an unsupported claim is what anecdotal evidence fallacy is all about.

I don't know Charlie's grandmother, I don't know what her health was, I've never heard of anyone smoking till 95 and being in good health. That is the correct defense, if the data given were "weak or poorly collected". If good data, the only defense is to enter wider more comprehensive data.

If you don't use that defense, the person giving the anecdote can say, 'The data refute it either, and it sure seems to refute the alternate claim.' Your point, that his point is wrong, because his data is incorrect, is lost.

" :Pseudoscientific data is unscientific data that is packaged to appear as scientific data. I don't see how this is relevant to the definition of anecdotal evidence.

Not anecdote.

It can be personal or drawn from worldly scientific study. It is the way the data was used not the data.

This is wrong. It is the origin (or lack of a verifiable origin) that determines if evidence is anecdotal. It has nothing to do with how the evidence is used. According to your definition, it would be possible to have two arguments based on the exact same evidence, but the evidence would only be anecdotal in the case of the fallacious argument. That's absurd! --Flatline 17:05, 3 January 2006 (UTC) "

Data is either scientific or false.

Packaging data to support a pseudoscientific claim, is called anecdotal, suppression of data, or false data, or unreliable data, depending on which flaw the pseudoscience is based on. Again, it is the claim that is pseudoscientific. Taking data pseudoscientifically produces false or unreliable data. Anecdotal evidence translates directly to short data or incomplete data.

The origin of a set of data determines its reliability not its anecdotal quality. If you measure gravity with poor equipment a calendar, you will get a poor value for "g". However, none of that information is anecdotal. If "Charlie's grandmother died at age 95...." the reliability is easily checked, but the use of it can be in anecdote or correctly. Note: It used to be that saying that proves not all smokers die of lung cancer was using the data anecdotally, but it has come to mean a scientific fallacy. The fallacy is one of support.

The strongest defense in that case is to show how the data is unreliable. It can still be an anecdotally incorrectly supported claim if the data is 100% accurate. The point is the correctness of the data raises the fallacy from poor data to another fallacy, that can be suppression of data, or unsupported anecdote.

The statement "anecdotal evidence" now implies the evidence is in error, but in reality, it used to imply that the evidence is short.

" :But not an orthogonal label. Any unverifiable anecdotal evidence would be considered poor data. Any anecdotal evidence, even readily verifiable, would be considered poor data if used to support a statistical conclusion since it is not framed in a meaningful statistical context. --Flatline 17:05, 3 January 2006 (UTC) " Again, it is the conclusion that is poor, not the evidence. If the evidence were poor, it would be false data.

" :Since when is it a fallacy to reference a relevant authority? Unless you have reason to doubt the authority of the reference or can demonstrate that there is no consensus of experts in the field, I am inclined to trust this reference. --Flatline 02:58, 5 January 2006 (UTC) " It is not a fallacy to reference a relevant authority. It is a fallacy to conclude that the reference to the relevant authority's conclusion, is proof that the conclusion is correct or even relevant.

You must present the authority's logic or data to prove your point. The fallacy was that you used his conclusion as a definitive statement. The debate here gains nothing. In fact I only have your word, here, that the "authorities" conclusion is relevant or a correct use of his words. I don't even know if he is the kind of authority you claim him to be. I also used that statement in punishment to the fact that you removed my earlier post.

Note: My earlier post just referenced your logical flaw by presenting logic and data. You are the one that chose to remove it.

Do a search on logical fallacies and look up Ipse dixit to get an alternate explanation to mine.

" :The data was incorrect. Your new article is similarly incorrect. --Flatline 02:58, 5 January 2006 (UTC) " The data was correct. My article is more correct than the one before it. It is evolving and improving. I'm thinking of adding the idea that anecdote is short story, hence, anecdotal evidence originally meant short data it has come to mean incomplete data, given to produce and erroneous conclusion.

" :I gave my reasoning on this very discussion page. In fact, you are responding to it now. --Flatline 02:58, 5 January 2006 (UTC) " If so I apologize.

It appeared to disappear, rather than become debated.

" :Please refrain from insults and name calling. It is inappropriate and distracts from the task of improving the article. --Flatline 02:58, 5 January 2006 (UTC) " Since you have now shown a willingness to respond to debate, I will delete the challenging personal parts of my posts.

--Eric Norby 09:57(PST), 2005, 29 December (AD)

The definition is incorrect.

Trying again.

Example: My grandmother smoked her whole life and lived to 96 without dying of lung cancer. That shows cigarettes don't cause cancer. Is an example of using anecdotal evidence. However, there is nothing scientifically nor statistically wrong with the data.

The only time a sample size of 1 is statistically valid is when you're disproving a conclusion that claims to be correct for 100% of the population in question (which really only happens in logical arguments, not statistical arguments). So, if you're trying to make the argument that cigarettes don't cause cancer, the argument is flawed because the statistical evidence you're arguing from is too weak to support the conclusion. However, the exact same anecdotal evidence could be used to support a very strong argument that cigarettes don't always cause cancer, so to say that an argument is fallacious just because the evidence it's based on is anecdotal is clearly wrong. --Flatline 17:05, 3 January 2006 (UTC)

So the definition that the data is often pseudoscientific or statistically unverified is incorrect. In fact, the evidence given can be shown to occur often, so the "singular" part of the definition not valid.

I agree that it is incorrect to say that anecdotal evidence is pseudoscientific. It would be more correct to say that anecdotal evidence is unscientific. However, generally speaking, it is correct to say that it is statistically unverified since if the anecdote were statistically verified, the argument would be based instead on the far stronger statistical evidence and not the anecdotal evidence. --Flatline 17:05, 3 January 2006 (UTC)

The fallacy is that the evidence doesn't support the conclusion. The conclusion that smoking doesn't cause cancer is erroneous. The correct statement would have been, Smoking doesn't always cause cancer many people including my grandmother have lived to 96 or so and not died of lung cancer while being smokers all their lives.

You're correct: the conclusion is too strong for the evidence given. But this does not support your argument that anecdotal evidence necessarily supports fallacious conclusions. --Flatline 17:05, 3 January 2006 (UTC)

Additional evidence showing that 90 percent of lung cancer cases are smokers relates the more relevant statistic. A large range of other experimental discoveries, and statistical, linking studies shows the mountain of proof that smoking is hard on your health. Science is about verifying things from a wide range of study, not one specific study, no matter how correct that study is.

At the worst anecdotal evidence is good solid data from a single study. All that was studied was those people living past age 90 that didn't die of cancer. Poor scientific conclusion. Perhaps poor study habits (Hand picking data, similar to data suppression but without intent.). But the data is absolutely correct and collected reliably.

If it weren't it would be false data.

It is a false dichotomy to claim that data can't be both anecdotal and false. --Flatline 17:05, 3 January 2006 (UTC)

If it weren't it would be pseudoscientific data.

Pseudoscientific data is unscientific data that is packged to appear as scientific data. I don't see how this is relevant to the definition of anecdotal evidence.

Not anecdote.

It can be personal or drawn from worldly scientific study. It is the way the data was used not the data.

This is wrong. It is the origin (or lack of a verifiable origin) that determines if evidence is anecdotal. It has nothing to do with how the evidence is used. According to your definition, it would be possible to have two arguments based on the exact same evidence, but the evidence would only be anecdotal in the case of the fallacious argument. That's absurd! --Flatline 17:05, 3 January 2006 (UTC)

Poor data has an entirely different label.

But not an orthogonal label. Any unverifiable anecdotal evidence would be considered poor data. Any anecdotal evidence, even readily verifiable, would be considered poor data if used to support a statistical conclusion since it is not framed in a meaningful statistical context. --Flatline 17:05, 3 January 2006 (UTC)

Anecdotal evidence is evidence used to falsely support a conclusion, given in an anecdote. The evidence given isn't wrong, the implied support is lacking.

--Eric Norby 2005 January 02, 10:10, (PST)

In television

In an episode of Dilbert, the cartoon show, Dogbert uses anecdotal evidence to persuade the public that they have work disabilities. Dilbert repetitively argues against the use of anecdotal evidence, but is proven wrong as Dogbert's evil plan unfurls and succeeds once again.

Sorry, as someone who has never (?) seen the show I don't understand this. Please rewrite this passage or preferably don't put it back into the article at all. <KF> 01:29, Dec 31, 2004 (UTC)

it does not follow that the conclusion is fallacious

This idea that anecdotal evidence only supports fallacious conclusions is wrong. Anecdotal evidence is simply evidence that comes from personal experience rather than designed scienfic experiments.

For example, if I know someone who was killed by a gun, then I can use that as anecdotal evidence that guns can kill people which is a verifiably correct conclusion even though it was arrived at through anecdotal evidence.

While it is true that anecdotal evidence is very weak evidence and is dangerous to depend on, this article goes too far when it claims that anecdotal evidence can only support an incorrect conclusion. --Flatline 17:29, 17 December 2005 (UTC)

The new article by Orzetto addresses my concerns. Thank you, Orzetto. --Flatline 14:32, 21 December 2005 (UTC)

Yes the conclusion is fallacious and the data correct.

The problems we are wrestling with on the definition of Anecdotal Evidence.

Truth: If the evidence is false, the fact that it is used in an anecdote is moot. The evidence is false, not anecdotal. It won't help to have poorly gathered data for the same reason, it is untrustworthy data, not anecdotal data. The data must be true or it supports nothing.

Personal: All evidence is observed from some personal position. It takes a person to observe data. Where the data comes from is unimportant to science. What is important to science is that it is verifiable and repeatable. The reason the data is often irrefutably true is because both sides of the anecdote often know the data is reported correctly from personal experiences. Drop a hammer on Earth and it will fall. It only needs to be true, not personal.

Singular: Often, but not always, users of anecdotal evidence quote more than one case. Therefore, singularity is not a required property of anecdotal evidence. Singularity is anecdotal evidence supporting the point that anecdotal evidence must be singular.

Anecdote: Evidence is science, the anecdote is what the evidence has been chosen to support.

Conclusion: The anecdote is the conclusion that the evidence was selected to support.

If the evidence supports a true conclusion, there is no surrounding evidence to prove the anecdotal nature. The conclusion then becomes a hypothesis or immature point, not an anecdote.

A scientific hypothesis takes data and attempts to fit a model to it. That data must be verifiable as is the data for an anecdote. The difference is, there is data that proves the anecdotal conclusion wrong, where the hypothesis remains a hypothesis until more data supports it or refutes it. If the hypothesis is claimed after refuting evidence is found, it becomes and anecdote supported by the earlier incomplete evidence.

If a conclusion is stated as verified by the data, the conclusion is immature. The cure for the immature conclusion is to point out additional evidence that could also explain the outcome.

Anecdotal evidence is an incorrect premise gained and supported by incomplete scientific data. The cure is to supply the complete data.

Premature conclusion is based on immature scientific data. Data gained without a full scientific investigation. Show additional possibilities.

The two are closely related but not the same.

The fallacy in anecdotal evidence is shown by the complete data.

The fallacy in the incomplete conclusion is shown by alternate explanations.

Smoking doesn't cause lung cancer, because many people have lived to 95 dying from other factors not relating to smoking or cancer, is anecdotal. The more complete evidence, that 90+ percent of the lung cancer cases were smokers, shows the fallacy. The anecdote exposed by more complete data, stops the necessity of more study.

Eating limes prevents scurvy, because Charlie ate them last voyage and didn't get scurvy, is a hypothesis or immature conclusion. There was no other supporting evidence to refute the claim. Being that it lacked falsifying data, more study was appropriate. Charlie ate a new thing he made and calls "aspirin" that voyage also, would lend addition avenues (Now both could be true.).

Quasars are powered, or not, by black holes, are premature conclusions. There is insufficient data to rule out either point. It is hard to know much about energy sources billions of light years away.

Hand picked data leads to false anecdotes.

Incomplete data leads to immature conclusions.

--Eric Norby 12:12, 26 December 2005 (UTC)

You seem to think that anecdotal evidence is some sort of logical fallacy and then go to great lengths to distinguish the "anecdotal evidence fallacy" from other logical fallacies. As such, the article you contributed, while containing some interesting points that should, perhaps, ultimately be included in the "final" article, is so heavily dependent on a misunderstanding that I decided to simply revert. Anecdotal evidence says nothing about the structure of the argument that uses it. Instead, it is a red flag that the premise might not be verifyable and that the argument using it might be relying on unsupportable generalizations (as my stats professor might say, "The plural of anecdote is not data"). Anecdotal evidence isn't a fallacy by itself, it's merely a catagory of evidence that hasn't had the rigorous safeguards of experimental evidence applied. Food for thought: should eye-witness accounts of crimes be considered anecdotal evidence? --Flatline 15:47, 27 December 2005 (UTC)

Okay, flipping through one of my philosophy textbooks from college (Invitation to Critical Thinking, by Joel Rudinow and Vincent Barry), in the chapter "Informal Fallacies of Evidence" they list a fallacy which they call the fallacy of suppressed evidence. From their discription, this fallacy matches the description that you've given for anecdotal evidence, except that in their description, the evidence can come from any kind of source (not just personal anecdotes). What's important to understand is that as a fallacy, this is a type of argument (as opposed to a type of evidence). Other fallacies that commonly use anecdotal evidence include: fallacy of small sample (where conclusions are drawn from a statistically insignificant set of samples), fallacy of unrepresentative sample (where samples are drawn from the wrong population), and (most obviously) unidentified experts ("I met doctor at the bar, and he said..."). --Flatline 17:49, 27 December 2005 (UTC)

"Okay, flipping through one of my philosophy textbooks from college (Invitation to Critical Thinking, by Joel Rudinow and Vincent Barry), in the chapter "Informal Fallacies of Evidence" they list a fallacy which they call the fallacy of suppressed evidence. From their discription, this fallacy matches the description that you've given for anecdotal evidence, except that in their description, the evidence can come from any kind of source (not just personal anecdotes)."

The above is a logical fallacy called "Ipse Dixit", "so his master says".

Since when is it a fallacy to reference a relevant authority? Unless you have reason to doubt the authority of the reference or can demonstrate that there is no consensus of experts in the field, I am inclined to trust this reference. --Flatline 02:58, 5 January 2006 (UTC)

• Remeoved Statement

The data was incorrect. Your new article is similarly incorrect. --Flatline 02:58, 5 January 2006 (UTC)

• Remeoved Statement

I gave my reasoning on this very discussion page. In fact, you are responding to it now. --Flatline 02:58, 5 January 2006 (UTC)

• Remeoved Statement

There is no reason for "Anecdotal Evidence" to be restricted to "personal anecdote". The better the data given the more it appears to support a fallacious anecdote. It is the logical leap to the conclusion where the anecdote errors, not in the data. The data is too incomplete to verify that conclusion.

If we accept your definition, then it is trivial to find argument pairs such that the data is anecdotal in one argument and not in the other. How can this be acceptable? Either the data is anecdotal, or it is not. The context is irrelevant; only the origin matters when determining if said evidence is anecdotal or not. --Flatline 02:58, 5 January 2006 (UTC)

• Remeoved Statement

--Eric Norby 09:57, 2005, 29 December (UTC)

Please refrain from insults and name calling. It is inappropriate and distracts from the task of improving the article. --Flatline 02:58, 5 January 2006 (UTC)

Rewrite

I had a go at a rewrite. The previous version, apart from being a hatchet job on those who use the term, contained a number of errors: particularly the claim that it isn't used in legal discourse. The idea that courts have some mechanism superior to science for checking witness testimony isn't true either: a search on "cognitive bias" + eyewitness will find various ongoing research that suggests the opposite. Courts still place a lot of reliance on unreliable cues such as witness confidence and detail of recollection. Tearlach 02:28, 19 January 2006 (UTC)

There is no category of evidence known in law "anecdotal". There is witness testimony and there is hearsay evidence, but not anecdotal. It was not claimed that the term "anecdotal" is never used in Court. It very likely is by medical expert witnesses for example, but it is not a category of evidence in law. Accordingly the above claim is wrong on that count.

There is no claim that law has superior or inferior methods to science. Law does however have methods for testing oral witness testimony which neither science nor medicine have or have developed. Again, therefore, your claims are incorrect. The statement that courts "place a lot of reliance on unreliable cues" indicates a misunderstanding of the function of a court. A court has an obligation to make a decision on the basis of the evidence before it, regardless of the fact that evidence from either side might be difficult to assess for reliability or might be of doubtful reliability. Further, by employing methods of testing evidence (which science and medicine lack) the reliability of evidence can be assessed. A court might decide a case by saying with reasons why one witnesses' evidence is "preferred" to another's if there is difficulty assessing the reliability of conflicting claims because of a lack of other sources to confirm or refute. Alternatively, there may be copious other sources to enable oral evidence to be assessed as thoroughly reliable or not.

Accordingly, it is clear from the foregoing that the prior edit was not a "hatchet job on those who use the term" but an accurate account of the position. Further it demonstrates that the "rewrite" is a POV job by someone who does not like what is said because it does not sit easily with his point of view. Additionally, User Tearlach pops up elsewhere with others who appear to be in a clique and engage in edit warrring with others whose views he and his colleagues disagree with.

Further, the use of "anecdotal" in relation to oral witness testimony is incorrect and misleading and that is a further reason why this "rewrite" is misleading.

Hence this misleading rewrite if left as it is will be yet another example of the kind of work with damages the reputation and value of Misplaced Pages and thereby leaves a big questionmark over any pretence there may be of being a reliable source of information. 86.10.231.219 07:26, 21 January 2006 (UTC)