Statistics and Cultural Illusions

[HydrogenBond]

Statisitcs are a useful tool in science. It allows science to proceed in areas that have not yet reached the level of rational understanding. In other words, if one can not logically predict results, one can run experiments in an attempt to correlate data, with the correlation being seed for theory. The hope is someday being able to find a more logical correlation that can predict data before experiments.

 

Statistical correlations are not fully logical, they are less than 2-D, i.e., cause and affect. They often have either a sound cause or affect, but the complementary second aspect of 2-D is usually fuzzy with some type of statisical and/or experimental uncertainty.

 

Let me give an example. If one did a study of the affects of one alcholic drink and reaction time and used 1000 test subjects, the result might go something like; it reduces reaction time by 20%, to just make up a number. This is a factual conclusion based on sound principles of math and at that level is 2-D, i.e., cause and affect of math. But on the other hand, this correlation is sort of noncausual, i.e., less than 2-D, with respect to a case by case study.

 

There will be people at the bottom end who, after one drink may lose 90% in their reaction time. There may also be those who drink all the time, feeling a little hungover during the study, whose reaction time may improve after one drink due to stabilizing their hangover, i.e, hair of the dog that bit you. Yet the study will present a one size fits all correlation that sort of reflects reality, at some level, but which is not quite in touch with causual reality at another level.

 

A good analogy is a man at a bar trying to pick up a girl. He may tell her his has money in the bank. If he does this with a cocky swagger, even if he only has $10 in the bank, he is telling the truth but the noncausual fuzziness allows room for spin. The hope is the female will spin the data herself and draw the desired erroneous conclusion.

 

In culture, statistical studies provide a true logical correlation in the strict mathematical and experimental sense, i.e, money in the bank. Because it is not globally applicable, i.e., rational correlation, there is a subjective element that is massaged socially by the spin doctors. The hope is that the money in the bank will be seen as being more than it is.

 

If the new hemoroid medicine Butt-out, presents studies of being 20% more affective, they hope you will beleive that money in the bank means 4 days instead of 5days, not the fact that it may not work on you and may cause some strange side affects. Statistical studies are ideally suited to sales pitch and spin but not logic.

 

The irony is that logic and common sense need to be avoided when it comes to statistical studies or it will neutralize the affects of the spin. But without the spin, the statistically studies lose half their power. Maybe it is a form of science entertainment that people love. Entertainment is more fun if one doesn't look to deeply into the special affects.

[sebbysteiny]

Statisitcs are a useful tool in science. It allows science to proceed in areas that have not yet reached the level of rational understanding. In other words, if one can not logically predict results, one can run experiments in an attempt to correlate data, with the correlation being seed for theory.

 

Not quite true. Statistics are more than simply useful, they are absolutely essential. It is agreed that under quantum mechanics, probabilities dominate outcomes and there is no possible way to predict which outcome there will be regardless of the information you know. The theory requires statistics and suggests that any other way of looking at the problem will produce an outcome incompatible with reality.

 

I admit this is a tangental point all but irrelivent to your point but out of loyalty to science, I think it's important to be accurate about these things :shrug:

 

But yes, statistics are essential in science. I know that this is a point that changes almost nothing about your arguments, but out of loyalty for science, I think it's important to be accurate on these things :).

 

Let me give an example. If one did a study of the affects of one alcholic drink and reaction time and used 1000 test subjects, the result might go something like; it reduces reaction time by 20%, to just make up a number. This is a factual conclusion based on sound principles of math and at that level is 2-D, i.e., cause and affect of math. But on the other hand, this correlation is sort of noncausual, i.e., less than 2-D, with respect to a case by case study.

 

There will be people at the bottom end who, after one drink may lose 90% in their reaction time. There may also be those who drink all the time, feeling a little hungover during the study, whose reaction time may improve after one drink due to stabilizing their hangover, i.e, hair of the dog that bit you. Yet the study will present a one size fits all correlation that sort of reflects reality, at some level, but which is not quite in touch with causual reality at another level.

 

Not quite true. Your above analysis will come from a simplified and false understanding of the statistics. In reality the survey showed a mean (20%) and a standard deviation. The distribution will almost certainly have a normal distribution shape (since all distributions tend towards a normal distribution shape on large numbers). There may also be a median and other statistical terms.

 

If you are given all this data then you can quite easily understand the link between alcahol and reaction times. All you need to do is work out your particular standard deviation and, by statistics, you could get a good idea of the effects. People who say "but I'm a heavyweight so I can drink slightly more than others" are doing just that: they are approximating their standard deviation and correcting accordingly and the results will probably be pretty accurate.

 

So the study will not give a 'one size fits all' answer. Media mis-reporting, however, might.

 

A good analogy is a man at a bar trying to pick up a girl. He may tell her his has money in the bank. If he does this with a cocky swagger, even if he only has $10 in the bank, he is telling the truth but the noncausual fuzziness allows room for spin. The hope is the female will spin the data herself and draw the desired erroneous conclusion.

You know I'm your greatest fan :D so I hope you don't mind if I say your pickup technique needs serious improving :(.

 

In culture, statistical studies provide a true logical correlation in the strict mathematical and experimental sense, i.e, money in the bank. Because it is not globally applicable, i.e., rational correlation, there is a subjective element that is massaged socially by the spin doctors. The hope is that the money in the bank will be seen as being more than it is.

 

If the new hemoroid medicine Butt-out, presents studies of being 20% more affective, they hope you will beleive that money in the bank means 4 days instead of 5days, not the fact that it may not work on you and may cause some strange side affects. Statistical studies are ideally suited to sales pitch and spin but not logic.

 

The irony is that logic and common sense need to be avoided when it comes to statistical studies or it will neutralize the affects of the spin. But without the spin, the statistically studies lose half their power. Maybe it is a form of science entertainment that people love. Entertainment is more fun if one doesn't look to deeply into the special affects.

 

It seems to me that your problem with statistics as a useful POLITICAL SCIENCE tool is NOT within the statistics itself but from the misuse and spin of statistics to make out they show something they do not.

 

From this, I agree one must be careful with statistics so that one does not leap to erroneos conclusions. But to eliminate their use altogether in finding truth even in political science is to be willfully blind to a very reliable wealth of evidence that acts as an accurate and effective tool exposing all truth including the 'science' of human political interaction.

 

The difference between us is you say that statistics can be spun. I say that statistics spun improperly are bad statistics. To properly use statistics one must become very familiar with statistical terms including standard deviation, mean, etc, and know exactly what they are. Then one must try and find the statistics that really matter including the values of all the above terms.

 

For example, a statistic that violent crime has gone down globally by 10%(used by Tony Blair to support his crime policy) is a much better statistic and getting understanding than the statistic that violent crime in one particular town (eg Grimsby) has gone up (used by Michael Howard to attack Tony Blair's policy). In my view, Tony Blair offered the fundamental statistic whilst Michael Howard offered spin only.

[Michaelangelica]

Could you please explain what you mean by "2-D"?

Thanks

[TheFaithfulStone]

Lies, damn lies, and statistics?

 

Of course, that's only statistically true, as only most and not all statistics are lies. :)

 

Statistics are useful during a lie, but by themselves are morally neutral.

 

And I second the 2D thing.

 

TFS

[Eclogite]

H-bond, Sebbysteiny has said it clearly. You appear to have completely misunderstood the character and function of statistics. What you are describing as limitations are limitations that arise when statistics are misapplied by the ignorant or the deceitful. That has damn all to do with statistics.

[Turtle]

By 2-D I mean something that is based on cause and affect. Cause and affect are the x,y axis on a 2-D grid, with logical relationships a curve on the grid, which always touch a combination of cause and affect. With statistics, either cause or affect are not absolute, but are fuzzy with some variability. This fuzzy data allows room for subjectivity when trying to draw curves, i.e, connecting bigger dots. Causual relationships use sharp little points that eliminate deviation from cause and affect.

 

 

Fuzzy like described here?

http://sipi.usc.edu/~kosko/IEEETFS2001.pdf

[Rebiu]

Sorry if I was misunderstood.

You were not misunderstood. You are wrong in your assessment.

But it does leave the door open to spin in the social and political arenas.

Only to those to who do not understand the nature of statistical analysis. In this light it no more opens the door to spin than any other analytical tool.

For the scientist statistics is a powerful tool for prediction where variability is of concern.

If applied correctly statistical analysis is useful not for prediction but for understanding what is happening.

But in the social arena this same variability can be used for spin.

No more than in any other arena or any other analytical tool.

With most people not truly knowledgeable of statistics, they become vulnerable to the spin.

This is true with the caveat that most people lack the capacity to deal with the complexity of contemporary issues and therefore make intuitive leaps as a substitute for careful analysis. The answer is not to illegitimize the methods that work. It is to educate the people.

The scientist may be immune, but that is less than 10% of the population.

This dichotomy of a scientist class in the population if you own personal fiction and has no basis in reality.

 

If the relationship is causal instead of statistical,

Another false dichotomy. There is no either/or aspect of causality and statistical analysis.

or solid math instead of statistical math, the average Joe, can plug into the equations and prove to themselves the ball will always fall down. Without statistics the spin doctors would be hard pressed to convince them otherwise. Statistics can be used as a tool or mind toy.

They seem to do just fine with tools like lying, exaggeration, quoting out of context, emotional appeal, ridicule and many more.

 

Statistics although very useful can inhibit the formation of causual relationships.

Causal relationships exist regardless of how they are analyzed. If you mean that the understanding of these relationships can be inhibited by statistical analysis I suggest that it is the erroneous application of the tool not the tool itself. If I try to drive a screw with a hammer it is not the hammer that is the problem when the board splits.

Statistics can do the job so well at times that one can assume this is the way of reality.

Statistical analysis is simple a way of processing complex data into a form that the human mind can use. Without the statistical analysis the information cannot be used. You are referring to an individual interpretation of this data. Without statistical analysis there is no way to interpret the thousands of data points that make up the raw information. Therefore you are advocating turning a blind eye to any information that requires statistical analysis to put it into a form that can be understood.

If one attempted to form a logical explanation, which may be closer to the truth,

Your illusionary dichotomy again. Statistical analysis does not provide explanations.

the statistics could created a self forfilling prophesy, that would subjectively dismiss the existance of a causual correlation.

The statistics do not create the "selfforfilling prophesy", the person that draws conclusions from the stats make that error.

For example, in quantum physics one assumes random variability to nature, with choas becoming whats in vogue. If someone tried to use logical considerations using unified force theory that shows the deviations are actually logical expected and not random, this would be met with pregidice since everyone assumes random, because the statistics work so well.

First of you reveal you ignorance about Quantum physics and chaos theory by applying the concept of random. Perhaps you could provide links as evidence that "everyone assumes random" as this is not the case. I have never heard the concept of random anything used in either of these theories. Eliminating the statistical analysis of the data in research of this nature would have brought the progress in these fields to a grinding halt long before this supposed split between theories occurred.

[Michaelangelica]

 

By 2-D I mean something that is based on cause and affect. Cause and affect are the x,y axis on a 2-D grid, with logical relationships a curve on the grid, which always touch a combination of cause and affect. With statistics, either cause or affect are not absolute, but are fuzzy with some variability. This fuzzy data allows room for subjectivity when trying to draw curves, i.e, connecting bigger dots. Causual relationships use sharp little points that eliminate deviation from cause and affect.

.

What sort of test of significance are you using/meaning? - correlation?

[Rebiu]

H-bond, Sebbysteiny has said it clearly. You appear to have completely misunderstood the character and function of statistics. What you are describing as limitations are limitations that arise when statistics are misapplied by the ignorant or the deceitful. That has damn all to do with statistics.

Well said. They clearly have no education in the theory and application of statistices.

[sebbysteiny]

For example, in quantum physics one assumes random variability to nature, with choas becoming whats in vogue. If someone tried to use logical considerations using unified force theory that shows the deviations are actually logical expected and not random, this would be met with pregidice since everyone assumes random, because the statistics work so well.

 

As Rebiu said, this shows no understanding of Quantum Mechanics. Einstein spent the later part of his life trying to eliminate statistics from quantum mechanics and 50 years later, the scientific community consideres this to have been a total waste. The statistics IS the causual relationship. It comes from wave particle duality and the heizenburg uncertainty principal. Particles with a 50% chance of being in one of two states neither behaves like a particle in 1 state nor a particle in the other. It behaves as if it is in both states at the same time until and only until it is measured.

[HydrogenBond]

Sorry if I was misunderstood. Statistics is very important tool in science. But it does leave the door open to spin in the social and political arenas. For the scientist statistics is a powerful tool for prediction where variability is of concern. But in the social arena this same variability can be used for spin. With most people not truly knowledgeable of statistics, they become vulnerable to the spin. The scientist may be immune, but that is less than 10% of the population.

 

By 2-D I mean something that is based on cause and affect. Cause and affect are the x,y axis on a 2-D grid, with logical relationships a curve on the grid, which always touch a combination of cause and affect. With statistics, either cause or affect are not absolute, but are fuzzy with some variability. This fuzzy data allows room for subjectivity when trying to draw curves, i.e, connecting bigger dots. Causual relationships use sharp little points that eliminate deviation from cause and affect.

 

For example, if one threw a ball in the air, it will fall back to the earth. This is a natural cause and affect relationship that anyone can see. If one collected statistical data, there may be a slight deviation added because of factors beyond control. The scientist realizes these slight deviation does not take away from the ball always falling down. The layperson, will see this this slight deviation and may assume in their imagination that the ball may sometimes, on very rare occations keep going up, since the relationship is not 100% perfect. The spin doctors can use this subjectivity and encourgage things that are out of touch with cause and affect.

 

If the relationship is causal instead of statistical, or solid math instead of statistical math, the average Joe, can plug into the equations and prove to themselves the ball will always fall down. Without statistics the spin doctors would be hard pressed to convince them otherwise. Statistics can be used as a tool or mind toy.

 

Statistics although very useful can inhibit the formation of causual relationships. Statistics can do the job so well at times that one can assume this is the way of reality. If one attempted to form a logical explanation, which may be closer to the truth, the statistics could created a self forfilling prophesy, that would subjectively dismiss the existance of a causual correlation. For example, in quantum physics one assumes random variability to nature, with choas becoming whats in vogue. If someone tried to use logical considerations using unified force theory that shows the deviations are actually logical expected and not random, this would be met with pregidice since everyone assumes random, because the statistics work so well.

[InfiniteNow]

the average Joe, can plug into the equations and prove to themselves the ball will always fall down. Without statistics the spin doctors would be hard pressed to convince them otherwise. Statistics can be used as a tool or mind toy.

Not if Joe threw the ball hard enough... hence, there is a probability somewhere that the ball will NOT return to the ground, and that can be represented through appropriate use of statistics.

 

As for cause and effect, nonlocality is gaining favor.