Which are more valuable? Real-life anecdotes or statistical studies?

[dustashes]

I think it really depends. Statistics are not necessarily unbiased and it is very difficult to pull a representative survey - even if it is representative, the ways questions are asked can have false results. Importantly, statistics can be used to persuade (another definition of statistics). For instance, there is a tendancy for political candidates to inflate support of their policies so as to draw more support. A healthy dose of skepticism towards statistics is necessary here.

Anecdotal evidence sometimes may be very useful, especially to highlight things missed out the majority (e.g. for minorities needs, rights, etc). For instance, to find out about medical malpractice (a hush-hush), anecdotal evidence can be used to find loopholes, etc.

 

[Ayatollah So]

It depends. Iuse both anecdotes and statistical studies. I don't feel one is inherently more useful than the other.

Good statistics and other sources of information inform and clarify each other.

I votes statistical analysis, but that needs a qualifier.

[...] statistics can be wrong. I therefore want to understand how their results came to be, and anecdotes can be a good source of information. If, e.g., anecdotes pile up like crazy saying a study is wrong, then that's a hint to have a close look at the methods of the study.

Which one you use depends on the circumstances. Sometimes statistics can [...] be conclusive but not give you enough useful information to make an informed decision.

I freely swap back and forth between quantitative and qualitative data. I learned the importance of nonstatistical data while working in market research for 10 years. Statistics tell us what the truth is. Anecdotes tell us how to understand it.

All of the above. Anecdotal evidence can clue you in to flaws in statistical studies that are far from obvious. It's not useful to ask "Suppose a statistical study is good - do you accept its conclusions?" because that question sweeps all the hard problems under the rug.

Further, carlos and Bestbank are onto something vital. Very often in research, right answers are obtained to wrong questions. For example, the eight-week effects of a drug may be studied, when it is common for the drug to be taken for a period of several years. The study may be flawless from a narrowly scientific viewpoint, yet not tell you what you need to know.

I still voted statistics, simply because at the outset of trying to learn about a subject, you're better off with a diet heavier in statistics than in anecdotes. But it's not either/or.

it is very difficult to pull a representative survey - even if it is representative, the ways questions are asked can have false results.

I'd say that the interpretation of the results is false, not the results themselves. Although, when the whole point of the poll was to create that false interpretation through deliberately weaselly worded questions, it can be a darn slim difference.

 

[Abegweit]

B) Statistical studies analyzing the issue using modern-day statistical methods with significant and conclusive* results (and unbiased funding of course)

There is no such thing as unbiased funding.

Do you trust your close ones and their anecdotes, or do you trust the statistical studies?

** - This is an example. Please discuss the original question. I don't want the thread to devolve into a discussion about vaccination, when it was meant as a throw-away example.

It's a very bad example. Anecdotes are facts. Theories need to explain facts. It is not difficult to explain why some people get diseases and others do not. It is far more difficult to decide whether, on balance, vaccines are good or bad for you. I really don't see how it is possible to do it statistically since studies, by their very nature, must examine one set of variables at the expense of others. Even more importantly, well-designed studies involving human beings are simply impossible. People are not molecules who mindlessly bounce off each other; they are thinking acting beings, something that statistical studies - by their very nature - cannot take into account.

Statistical studies are not theories. They explain nothing; at best they are the equivalent of anecdotes - facts that need to be explained by theory. Far too often, they are nothing more than rigged experiments designed to provide the answer that the experimenter and his backers want.

Today science is pretty much dead, theory has been replaced with biased studies which purport to show something. Dietary science is perhaps the best example of this. Studies contradict each other repeatedly and we are supposed to pretend that they prove something when manifestly they do not.

The other favourite tool of modern quackery is the computer model. If anything, models are even worse than statistical studies since they are programmed to come up with the designed answer. Rigging is an inherent part of the process. It cannot be avoided. Climate and economics are best examples of this particular form of pseudo-science fakery. For example, the government "economists" around Ben Bernanke that, had they not printed trillions of dollars and given them to their cronies, the unemployment rate would have been 2.5% higher than it actually was. Nothing can prove this wrong. Pointing out the fallacious nature of the theory on which the models are based simply does not work because the PTB have a vested interest in the results.

So I will take anecdotes over either studies or models. Anecdotes are facts. A study may possibly be a fact as well, but it is inherently suspect. Even if the study actually uncovered a fact, a good theory is required to explain how these facts fit together. Without theory to back them up facts are nothing.

Studies and models are substitutes for science whose purpose is to prove that the correctness of particular point of view, usually some PC nonsense which contradicts common sense. They do not advance science in any way.

 

[SS-18 ICBM]

People are not molecules who mindlessly bounce off each other; they are thinking acting beings, something that statistical studies - by their very nature - cannot take into account.

Wow, you do not even know what you are talking about. And worse, you have the the chutzpah to be smug about it too. Ever heard of a control group?

They do not advance science in any way.

Really? I guess HAART is completely ineffective against AIDS, seeing as the theory behind it requires modeling of HIV's evolutionary response to antiretrovirals and statistical studies of patient responses and viral loads.

 

[Defiant47]

There is no such thing as unbiased funding.

Yes there is, unless you're a massive conspiracy theorist.

It's a very bad example. Anecdotes are facts. Theories need to explain facts.

Anecdotes are single facts. Statistics analyze the facts over an entire population.

It is not difficult to explain why some people get diseases and others do not. It is far more difficult to decide whether, on balance, vaccines are good or bad for you. I really don't see how it is possible to do it statistically since studies, by their very nature, must examine one set of variables at the expense of others.

Actually, that's why we have regression analysis.

Unless you're talking about the whole "analyzing the population is impossible with just a statistically significant sample", in which case you don't understand statistics.

Even more importantly, well-designed studies involving human beings are simply impossible. People are not molecules who mindlessly bounce off each other; they are thinking acting beings, something that statistical studies - by their very nature - cannot take into account.

Many times that is irrelevant, or covered by some sort of control group.

It feels like you're demonizing a science simply because it focuses on one aspect of human beings (i.e. will this thing make you sick?), and thus doesn't take into account feelings.

Statistical studies are not theories. They explain nothing; at best they are the equivalent of anecdotes - facts that need to be explained by theory. Far too often, they are nothing more than rigged experiments designed to provide the answer that the experimenter and his backers want.

They're better than theories. They're conclusions. And they explain everything, while anecdotes explain one thing.

Statistics don't lie. Humans lie. And there are ways to misuse statistics to get the results you want, but most of them are easily caught by anyone with a basic understanding of statistics. The fact that something can be misused by bad people, does not discredit the entire field.

Today science is pretty much dead, theory has been replaced with biased studies which purport to show something. Dietary science is perhaps the best example of this. Studies contradict each other repeatedly and we are supposed to pretend that they prove something when manifestly they do not.

The other favourite tool of modern quackery is the computer model. If anything, models are even worse than statistical studies since they are programmed to come up with the designed answer. Rigging is an inherent part of the process. It cannot be avoided. Climate and economics are best examples of this particular form of pseudo-science fakery. For example, the government "economists" around Ben Bernanke that, had they not printed trillions of dollars and given them to their cronies, the unemployment rate would have been 2.5% higher than it actually was. Nothing can prove this wrong. Pointing out the fallacious nature of the theory on which the models are based simply does not work because the PTB have a vested interest in the results.

So I will take anecdotes over either studies or models. Anecdotes are facts. A study may possibly be a fact as well, but it is inherently suspect. Even if the study actually uncovered a fact, a good theory is required to explain how these facts fit together. Without theory to back them up facts are nothing.

Studies and models are substitutes for science whose purpose is to prove that the correctness of particular point of view, usually some PC nonsense which contradicts common sense. They do not advance science in any way.

Yeah, actually, I'm pretty sure all I have left to say is that you have no idea what you're talking about. Or rather, you think you do...

 

[Catharsis]

In most cases, when forming an opinion I would use anecdotes more readily than statistics, but that's just because anecdotes are easier to come by. It's not by any means the correct thing to do. It's just that I'm lazy.

Honestly I do not understand the suggestion implicit in the OP's question that statistical studies do not take place in 'real life'.

If you really want to influence people, you compile your stats and then make up a story about the results. You create a narrative that makes the data anecdotal and interesting.

Well sure! There's all sorts of graphics and tables and such that can make the stats easier to digest if that's your thing. It's still the result of a study, though, and it shouldn't be jettisoned because of that, in favour of Dave down the pub who swears blind that the MMR jab gave his littlest autism.

 

[Abegweit]

Wow, you do not even know what you are talking about. And worse, you have the the chutzpah to be smug about it too. Ever heard of a control group?

Yes, I've heard of control groups. Certainly studies with control groups are better than observational studies but they are no sinecure either. There are plenty of examples of controlled studies which have been failures and deliberately-designed failures at that. FYI, I understand statistics and also I understand far too well how they can be abused.

As for Chutzpah, the arrogance of people who have spent many years learning how to replace thinking with studies and models is quite simply breath-taking and, you manifestly are a prime example of those who have done so.

The most offensive part of this whole scam is the way that these so-called scientists mix up observational studies with controlled studies and are quite happy to let the ignoramuses in the press draw invalid conclusions from their manifestly deliberate conflation of the two. On top of that, they all too frequently pass off correlations without any statistical significance whatsoever as scientific fact. The fact that they are drawn from badly-designed studies is just the icing on the cake

The nonsense about second-hand smoking is an excellent example. There is absolutely no evidence whatsoever that second-hand smoking has any effects on health at all. The approach to this issue was scatter-shot. Compare non-smokers who live with smokers and try find ANY correlation with ANY disease. Eventually you are sure to find something. And if the p-value is absurdly low, well who cares? The purpose of the study is to prove the pre-desired outcome. As such something which is badly designed which uses poor stats is clearly preferable to honest work. The fact that such garbage gets published in places like The New England Journal of Medicine is proof of the extent to which science has been turned into a tool of politics. For the record, I am not a smoker. I quit many years ago.

As for HAART, I know nothing about it. As such I cannot judge. I will say that any scientific claim which uses the word "model" is immediately suspect. But perhaps this is an exception.

@Defiant47 Your touching belief in the purity of certain funding sources truly warms my heart. Still... you haven't explained what it is that magically leads one source to be honest and unbiased and what it is that leads another to be suspect and rejected simply because it's tainted. Personally I don't care where the funding comes from. I analyse results based on whether they actually appear to be based on science - e.g. on theory which attempts to provide a explanation for how the real world works and especially if the theory appears to actually make sense. If the scientist proposes methods to disprove his theory, he gets bonus points. Unfortunately those cases are few and far between.

I do have a few rules about estimating scientific claims which I haven't actually examined properly. Here's a few

1) Any sentence that starts with "studies show..." is not science and should rejected out of hand. As I said, science is about theories. Studies, in themselves, are worthless. More generally, all facts are worthless in themselves. Science is about theory which weaves facts together. The main difference between anecdote and study in this respect is that the former is less amenable to manipulation.

2) The PC side is almost always lying and the other side is usually (but not always) telling the truth. There are many reasons why this is true but possibly the best is that the press always accepts PC conclusions at face value. In contrast, the other side has to actually defend its claims. In particular, the press will go out of its way to find PC experts to deny the un-PC theory. Said expert doesn't actually have to refute the un-PC theory; he simply has to deny it.

3) Most computer models are lies. Now there is an important distinction to be made between models which attempt to accurately reflect reality (say, protein folding) and those which simplify it (say, CO2 forcing). An even more important distinction exists between those which are designed to reinforce PC claims and those for which neither side is PC. Again protein folding vs. climate forcing is a pretty good illustration.

4) Anyone who doesn't propose a theory to back up his analysis is at best a hack and usually an outright liar. In any case, he certainly is not a scientist.

 

[SS-18 ICBM]

Cite me enough studies that are examples of what you said. Quantify your assertions.

The main difference between anecdote and study in this respect is that the former is less amenable to manipulation.

Ahahahahahahahahahahaha, oh wow!

 

[GoodGame]

3) Most computer models are lies. Now there is an important distinction to be made between models which attempt to accurately reflect reality (say, protein folding) and those which simplify it (say, CO2 forcing). An even more important distinction exists between those which are designed to reinforce PC claims and those for which neither side is PC. Again protein folding vs. climate forcing is a pretty good illustration.
.

Regression is lies!!!! All lies!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!

I guess the alternative is better because it feels nicer when its graphite on cellulose?

I'd suggest learning some more about stats and regression analysis, etc..

My useless generalizaton is: no model is guaranteed to be perfect, but that doesn't mean they don't have strong and measurable predictive power.

 

[Abegweit]

Cite me enough studies that are examples of what you said. Quantify your assertions.



Ahahahahahahahahahahaha, oh wow!

Such an awesome argument! I bow to your cognitive powers and apologise for daring to bring logic into the discussion. Obviously "Ahahahahahahahahahahaha, oh wow! " trumps any attempt to actually think.

 

[Abegweit]

Regression is lies!!!! All lies!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!

I guess the alternative is better because it feels nicer when its graphite on cellulose?

I'd suggest learning some more about stats and regression analysis, etc..

My useless generalizaton is: no model is guaranteed to be perfect, but that doesn't mean they don't have strong and measurable predictive power.

My [not so useless] generalisation: arrogant put-downs are far easier than actual attempts to come to terms with a point.

I suggest learning some more about damned lies and how they are a cut above "stats and regression analysis, etc.."

Second [not so useless] generalization: those who defend models have a lot of [so-called] work invested in proving the value of damned lies.

 

[Millman]

A personal preference depending on the most exact nature of the situation.

 

[SS-18 ICBM]

Such an awesome argument! I bow to your cognitive powers and apologise for daring to bring logic into the discussion. Obviously "Ahahahahahahahahahahaha, oh wow! " trumps any attempt to actually think.

I don't need to. You were obviously ignorant of how important those "useless" models and statistics are to modern pharmaceuticals. And you persist in glorying in your lack of knowledge. There is no point arguing here.

 

[Bill3000]

wrong thread

 

[Abegweit]

I don't need to. You were obviously ignorant of how important those "useless" models and statistics are to modern pharmaceuticals. And you persist in glorying in your lack of knowledge. There is no point arguing here.

Wow. Just simply basking in your own glory. No need to prove anything or even to defend it. Simple proof. You are God and, as such, your pronouncements are clearly right. I apologize for even thinking that such a simple minded notion as logic could contradict your Magnificient Being.

 

[Tee Kay]

SS-18 ICBM is a science expert around here.

 

[Dachs]

Wow. Just simply basking in your own glory. No need to prove anything or even to defend it. Simple proof. You are God and, as such, your pronouncements are clearly right. I apologize for even thinking that such a simple minded notion as logic could contradict your Magnificient Being.

I don't see where the logic came into it. Speaking from a historical perspective, the entire concept of historiography (or at least half of it) is devoted to figuring out wie es eigentlich gewesen ist ("how things actually were") despite the unreliability and inaccuracy of sources, who all have their own skewing points of view, if not agendas. Anecdotal evidence in the form of historical writings are very much open to manipulation. I doubt you could quantify it as "more" or "less" open to manipulation than statistical findings (a skewed presentation of subject matter, incomplete sets, or whatever). I think you're getting lost in the purely scientific aspect of things and in a few selected real-world situations and not talking about the actual question in the OP.

 

[SalmonSoil]

Heres the thing, statistics are facts. Anecdotes are only facts concerning the events in the anecdote (and sometmes not even then). Yes statistics can lie, and be misused. But anecdotes lie worse because they imply the situation is true for the entire group, and anecdotes are frequently misused because it is far easier to come by an anecdote supporting your fallicy than it is to find statistics. Where ever statistics are biased, anecdotes are doubly so.

 

[Xeus]

When you seek to form your opinion about a subject, which has the greater pull on you?

Do you trust your close ones and their anecdotes, or do you trust the statistical studies?

This is too simplistic - too either/or.

I personnally form my opions based on a wide range of data AND anecdotes. And unlike many people who, once having formed an opinion, who will stick with it and defend it even in the face of overwhelming contrary opion or evidence, my opinions are organic - they are modified according to new anecdotes or statistical/scientific evidence.

In regards to anecdotes - some people I'm really close to can also be very stupid (but I love them very much), so it pays to listen to them "with a grain of salt". Also, just because a lot of people are saying the same thing dosn't make them right (e.g. flat earth, God, global warming, UFO's, etc. [Just examples- I'm not discussing them]).

Now, in regards to statistics, there are many problems. Firstly, you have to understand the discipline of statistical analysis, in which I tried and basically failed, I know just enough to make me dangerous. We are all familier with the 'lies, damed lies and statistics' line. I always ask myself these questions:
- exactly what questions are being asked?
- has a broad, say socio-economic/gender/age/ethnic group been studied?
- Is there a blind study?
-who is asking them?
-do they have an agenda?
-who pays their wages?
-how large is the study?
-what policies or studies could be further implemented from the results?
- will the results be published in a scientific peer-reviewed journal, or is it in Gangsta Rap Quarterly.
etc, etc.

I think (I hope!) most people weigh up the ancedotal and statistical evidences and probably form an opinion - often the middle ground between two opposing views. I think most people apply (if they don't allow their emotions to get carried away) what is generally referred to as "common-sense".

 

[Brighteye]

The most common statistical error is to find a significant effect (statistically) and equate that with significant (general usage), which implies a large and meaningful effect.

When you see someone reduce something from 100 to 90% and someone else say that this something has been blocked, you can easily be fooled into thinking that the effect is total.