Statistics and reason

[pioneer]

I am not trying to undermine the practical usefulness of statistics, when used properly by a trained mathematician. But from the practical point of view of the lay person, a statistical approach can get irrational, since it can be used to manipulate a reasonable lay person.

 

Let me give an example. Say eating X, doubles the risk of a person getting condition Y. The math expert has a good handle on this in terms of the hard numbers. It can tell the difference between going from 1 to 2 in a million and 100-1000 in a million. But the layperson is pitched "double the risk", for the 1 to 2 in a million. They are not experts so they then depend on others to interpret this danger. If I want to sell product Z, that halves the risk, I am not going to be clear about what this all means. I am better off pitching risk angle, where competitor X is double, which is really terrible. I will market myself offering the customer half risk. The math expert will see no real difference between the products. But the housewife is now going to buy brand Z with the irrational fantasy she just saved her family.

 

With reason one can not successfully create the same impact. There is truth in advertising that does not allow one to say x=y, if x=z. But with statistics we can get around this by providing a mathematical truth that is misleading. The double the risk is indeed true. I just have to my pitch product in a way were I don't technically violate reason. The math expert will see the game, but how many experts are there? Why is there no voice of reason sounding out? One possibility, is the experts are also living in a world that is not fully rational, because it assumes a world of chaos with odds allowing almost anything. The misleading pitch fits into this chaos, since there are the odds they are telling the truth.

 

Is it possible that chaos is an pre-reason affect and not a rational cause? In other words, if x=y, but we don't know this, then doesn't than mean that anything is possible, with some things more possible than others. With that assumption then x can never exactly equal y, or it can never reach reason, with the slight difference equal to the laws of chaos. The math expert can see the proper interpretation of the applications but can it see the possibility of a pre-rational cause and affect, mix-up. It is the chicken or the egg. Once you decide, there are two paths that work. One may put cause before affect and be able to reason since it will follow from there. The other may have to put the affect (chaos) before the cause. Or chaos will affect cause so it can lead to statistical type affect, which lead to further chaos.

[the tree]

You, and basically everyone one in the world should read How To Lie With Statistics by Darrell Huff. The contents is simple and most of it could be understood by the lay person.

 

By the Freedom Of Information act, you can obtain data used to make outlandish statistical statements, and if you are so educated you can do the hypothesis testing yourself.

 

There are plenty of independent organisations, people, websites etc etc who are dedicated to taking a critical look at "scientific" claims. Ben Goldacre being a prime example of this.

 

I have no idea what your last paragraph is on about.

[pioneer]

Addendum

 

I have no idea what your last paragraph is on about.

 

Statistics is a powerful math tool. Chaos is one of its performance parameters that allow it to work. The parameters of the tool does not mean this is how reality is, only how the tool needs to view reality to work properly. it sort of like driving at car that can go 120 mph and then assuming all cars have this same speed parameter. If one only drives that car, and never tries another, this assumption can start to look like it is reality. If others start to believe that limit is true, they may never try to exceed 120 mph, since that is now the established limit. If that is the limit then we only build car with that limit; the loop closes until the line between cause and affect becomes a vague memory.

 

I am trying to walk down the middle since statistics is a very powerful tool. But I am also trying to show how chaos is part of the technique, and can be used to manipulate reason.

 

The reason we use would prefer to use statistics, in the first choice, is when systems get too complicated for reason. That is why it is used in medicine so much. Since the system is too complicated for reason, we need to work under the parameters of the tool, such that anything is possible, since there is no reason, it is not. Based on that assumption we let the data and the math give us a correlation. This is totally valid and very practical with plenty of very useful results.

 

But what we have seem to have forgotten is this approach was needed when reason started to break down due to increasing complexity and that it started out its young life as a rational approximation to help reason. It was very ingenious allowing science and production to move forward. The wide range of success has sort of created the situation where the reasonable is being taken over by the approximation. Even if x=y, we now need to assume chaos, as part of the reasoning, to make it an even more complicated system. Chaos is not needed at this low level of rational complexity, but its value for complex systems seems to logically imply it can do the easy ones, too, even better. There is less incentive to reason in areas of science, but to let chaos do it for you.

 

Here is an analogy. The workers are tilling the soil of science. The come to this huge bolder in the field. All the humans can't lift it, so we call in the backhoe. The backhoes picks it up, with reasonable dexterity, and places it where we need it to be. Soon, since the backhoes can do what no human can do, maybe it is also better at reasonable tasks like picking up the pebbles and rocks There is always going to be greater uncertainty in backhoe placement of the object. The backhoes can get a lot closer with the boulder since the percentage is smaller. The pebble has huge placement uncertainty as a percentage.

 

If we assume the placement uncertainty of the backhoe is the new standard even for pebbles, then the power of reason seems off, since it can place the pebble in the bull eyes. This is not suppose to happen, using the backhoe and chaos. So what happens next, humans with the pebbles and rocks begin to reason from a distance, so that can toss the rocks to meet the prescribed uncertainty. There is no reason to hit the bulls eye. If I hit the outer ring and add a couple of quickie chaos's, it will now considered a bulls eye. This will reach the lay person like it is based on a rational bull's eye. Reason has a certainty of 1 which is not possible with chaos . So in chaos rules, it is impossible to reason, since it assumes 1.

[the tree]

Statistics is a powerful math tool. Chaos is one of its performance parameters that allow it to work. The parameters of the tool does not mean this is how reality is, only how the tool needs to view reality to work properly. it sort of like driving at car that can go 120 mph and then assuming all cars have this same speed parameter. If one only drives that car, and never tries another, this assumption can start to look like it is reality. If others start to believe that limit is true, they may never try to exceed 120 mph, since that is now the established limit. the line between cause and affect becomes a vague memory.

Th If that is the limit then we only build car with that limit; the loop closes untile thing is that statistics allow you to say things with a certain level of certainty, a sample space of 1 allows almost no certainty. There is such a thing as good statistic and crappy statistics. What your talking about is making an assertion based on not enough evidence, this is a symptom of human folly and nothing intrinsic to statistical theory.

 

As a point of semantics, chaos is not synonymous with unpredictability or haphazardness. Statistical approaches are okay with certain degrees of randomness or *gasp* indeterminacy, but chaos is something very specific that statistics is quite happy without.

[Bignose]

"The are lies, damn lies, and statistics" -- Mark Twain

 

Basically, what you are looking for is some actual practical science to be taught in the classroom. Unfortunately, this just isn't the status today. The kids are taught science as just another collection of facts to memorize. The colors in the spectrum are ROY G. BIV. The eras of the earth are Cenozoic, Mesozoic, Paleozoic, etc. The distance from the earth to the moon is 384000 km. Etc. Etc. Etc.

 

It is even to the point where the steps in the "scientific method" are just a list of things to memorize. There is no practical discussion/use made of them.

 

There is no skepticism/burden of proof taught in schools. That's why Head On as a product can exist for more than 5 mins. If everyone learned some basic skepticism, everyone would see right through the claims. Same thing with the magnetic bracelets.

 

The statistics come into play too, because sure these products claim some improvement/healing... which is true. But, the improvement isn't statistically significant above placebo. I'd like to see them spend an entire year of science classes just on placebo & statistical significant if that is what it took to teach people about this.

 

Part of it also is that large numbers are essentially meaningless. A guy last week got two holes-in-one in one round. Golf Digest said that the odds were 1 in 161 million. What does that really mean? Why even quote it (other than it makes a catchy sound-bite-like phrase)? Why the accuracy to 3 figures -- 1 in 100 million or 200 million is practically equivalent.

 

The numbers are especially troublesome when health reports come out. Especially when they report the relative statistics. "Eating bacon increases your chance of toe cancer 35%" (Please note I completely made that up.) Well, if you chances of toe cancer were 1 in 100,000, they are now 1.35 in 100,000 if you eat bacon. Not a large increase, especially if you like to eat bacon. But, saying 35% makes it look like a huge dramatic increase in risk. Same thing about the health benefits -- "drinking coffee lowers the risk 10%" Again, if it was 1 in 100,000, if you drink coffee, now it's 0.9 in 100,000, again pretty tiny. And, if you don't like coffee, then it almost certainly isn't worth forcing it down.

 

Dr. Dean Adell has a very good book that expounds on this: Eat, Drink, and be Merry. He talks about how the media likes the relative statistics because it makes it seem more significant. But, the basic tenets of good health have been known for a long time -- eat a wide variety of things and exercise more often than never and your chances of a long healthy life are pretty good.

[pioneer]

That is part of the problem I see. If someone misquoted the distance to the sun, there would be someone to correct it. If we misuse statistics for fun and profit, there is nobody correcting the record to keep it honest. Those knowledgeable in these areas have an immunity but they don't try to pass that on. It may be something simple like conflict of interest. Even if this is being misused it still creates jobs in the field.

 

One area that is really abused is polling. Each team gets their experts to run the data they need to get the result that they want. It a lot of it has to do with spin. That is why it is not uncommon to have it both ways. These, results in turn, can change day to day. The impression I often get is that it is being used to stir the herd with the power of suggestion, but using the label of scientific so it appears like science.. It could also be used by the media to estimate how effectively they are manipulating their target audience. The goal is see if their impact is shifting the direction of the herd. I am not down on the actual math when used in the hands of the expert, but there is no other math that can be used as well for manipulation.

[Bignose]

I dunno, pioneer. What "tool" hasn't ever been manipulated by those with a motive? Math/statistics/science is only one entry in a long list of things that people use to sound authoritative with, but is really just another way to try to grab power. State-run media (like in Russia, Zimbabwe, Myanmar) is another. Religion is probably the real biggie. How many things have been done in the name of a god?

 

It is just another thing that needs to be viewed with skepticism. It is easy to just look at a statistic like "red headed people get 8% of the tickets, but are only 4% of the population, so they are being profiled unfairly" and agree with the conclusion that is made. (** Just to be perfectly clear, I completely made that up**) But, the real question is rarely so simply clear cut. A fair number of red headed people have some Irish lineage, and there are a fair number in New England maybe the New England police just write more tickets than the rest of the country and it appears that the red haired people are being targeted. What is the sample size? which leads to the question what is the chance that it is just coincidence?

 

There was a significant apple scare in the 1970's -- a medical study linked cancer and apples. As you can image, for a time the sales of apples completely plummeted. But, it was all a coincidence. The subjects of the study just happened to get caner at a higher rate completely naturally. And, this was only one study, out of many others that showed the benefits of apples -- but the media ran with it -- and like I said, the sales of apples plummeted. It didn't take too long to realize what happened -- but it is important to note that it wasn't a mistake in the traditional sense -- there was a strong correlation between the subjects of that study who all ate apples and cancer. But, correlation does not mean causation.

 

Anyway, like I said, I think that statistics is just one more thing to manipulate today. There was a time -- late 90's -- where you could predict the outcome of a global warming study with pretty good accuracy just by looking at the funding source. You can still accomplish that today with gun studies.

 

It really doesn't matter who is telling you anything -- the important thing is to use your own brain and don't just take someone's word for it. Ask for some evidence to back things up. Ask the tough questions. Don't let someone get away with using any of the common logical fallacies like appeal to authority. Don't let someone turn it around on you just for asking questions "How do you know what I say isn't true" -- it doesn't work that way, the claimer has to show the evidence for his statements, you don't have to entertain any and all whimsical notions until they have been disproven. Skepticism today seems to have become almost a dirty notion. I've seen people become almost angry at me for asking for evidence of the statements they made. As in political correctness taken to some extreme says that we have to accept all different ideas until proven wrong. But, that just isn't right -- if someone makes a statement, it isn't an unreasonable or unfair or wrong point of view to ask them for some evidence to back up their statements. I'm not talking about opinions, but when they make factual-like statements. Skepticism is something there is precious little of today, so just always look at these things through a skeptical eye. Stats are just another thing used to skew points of view, and I think that they need to be viewed as such.

[foodchain]

Wow, this thread is refreshing. What would happen if all of math and its uses were only done in statistics, what do you think human understanding would look like? I mean could you model gravity with just stats?

[PhDP]

Wow, this thread is refreshing. What would happen if all of math and its uses were only done in statistics, what do you think human understanding would look like? I mean could you model gravity with just stats?

 

You can (in fact, must) use statistics to test hypotheses, and you can use statistical inductions to uncover relationships.

 

But you can't have a theory of gravity based solely on statistics. Gravity does not depend on probability (it's not a stochastic procress), at least, not according to the theory of relativity... As such, there's no point to model gravity with just stats. And anyway, even statistical physics (used, for example, to study thermodynamics) depend on much more than just statistics, it relies on algebra, calculus, and probability theory (which is an important foundation of statistics, but not considered within the field of statistics).

 

There was a significant apple scare in the 1970's -- a medical study linked cancer and apples. As you can image, for a time the sales of apples completely plummeted. But, it was all a coincidence.

 

We live in a society fundamentally hostile to the concept of 'coicidence'. You're the lone survivors of a plane crash ? It's fate ! or god ! or destiny (of course, the other ones were just bad people)...

[Bignose]

We live in a society fundamentally hostile to the concept of 'coicidence'. You're the lone survivors of a plane crash ? It's fate ! or god ! or destiny (of course, the other ones were just bad people)...

 

That's very true. To borrow a statistic from baseball the difference between a 0.300 hitter (not awesome, but good enough that in the major leagues you're going to keep your job) and 0.250 hitter (perceived as below average, needing improvement) is one hit a week. One extra hit in an average of 6 games a week.

 

But, a player with a .250 average is viewed as "poor" on a major league level. That the guy needs some help. It may just be that he's had some hard luck -- his mechanics are solid, and he's hitting the ball well, he's just hitting them at the fielders. The major league hitters can hit it to different fields, but no one is skilled enough to hit major league pitching to very specific spots -- and a guy hitting .250 for a time, may be in fact be an actual .300 hitter, but just had some bad luck. Even top of the game players like Alex Rodriguez or Albert Pujols will have stretches where they hit 0.250 for a month and nothing is wrong.

 

It is human nature to want to look for causations for things -- it is fundamental to mankind's thirst for knowledge. But, in that quest, we have learned that some things aren't always going to have direct causations. Some things are going to behave statistically/probabilistically. Quantum mechanics is probably the biggest one, but things like statistical mechanics of gases, turbulence in fluid flows, the stock market, the agglomeration/flocculation of particles or oil droplets in fluids, Brownian motion, birth-death processes in bacteria/cells, etc. etc., all have very high dependence on random variables as near as we can tell today.

 

Like I said above, I would not have a problem with science classes taking an entire year -- maybe about 8th grade or even 10th grade -- and driving home some of these issues. Teach skepticism and learn how to interpret the magnitude of the numbers that are reported. Teach how correlation does not imply causation (I really like the example of how ice cream sales and shark attacks are very highly correlated in the U.S. and that only implies causation if you want to say that the sharks like to attack people who eat a lot of ice cream). Teach just how often coincidence really and truly does occur. (A good example is my fiance had a period of time when she was in 6 car wrecks in 6 years, none of which were her fault -- I said "well, that just shows that it had to happen to someone. The entire curve gets filled out with several hundred million drivers in the U.S., and there are enough accidents that it just happened to happen to you. Someone out there has had 10 wrecks in 10 years, so be happy that isn't you" Like I said, every time, the police determined it wasn't her fault, so it is nothing but pure coincidence. She isn't driving poorly to get into all those wrecks, she just was in the wrong spot at the wrong time fairly often.)

 

I don't know if a class like that would help, but there are some very fundamental misunderstandings out there today that I think could be remedied reasonably easily if an effort was made to remedy them.