Sometime ago I had a conversation with an elderly friend of mine and I was told that until the mid sixties no one locked their house. I find such thing quite difficult to imagine in the today's society, especially after all the efforts of politicians, humanists, sociologists, implementing the state of the art democracy. I enquired from others who remember that time and they confirmed that back then people indeed did not lock their houses. Thus by the testimony of several most definitely trustworthy elderly people, from three quite distinct European countries with very little interaction one with the other I accept for established fact that until the mid-late sixties people did not lock their houses and cars. Further it also seems that in the past people were happier and healthier. The later might contradict what the media usually states, but one could simply look at the life span of people in an old graveyard or the life span of famous scientists, artists, musicians, etc. from the past and make his/her own mind. It seems that if a person had a relatively decent lifestyle (a counter example is Schubert who died from syphilis at age of 31) or hadn't had some unfortunate accident due which they would prematurely die, most people would have lived to their seventies and later. So if society is getting better as the politicians and media claim why society is obviously getting worse? Similarly, if there are always new and better drugs – why people are getting sicker? In this article I will look at one of the reasons for this misunderstanding. But first let us look at some definitions, namely what science is:

I claim that statistics and probability (from which statistics derives number of concepts) are unscientific self serving methodologies yielding meaningless and unattached to objective reality results. As such these results can be (and usually are) exploited in ways the party using statistics and probability finds fittest for achieving its own particular objectives. Which explains the above observation that what people are told does not match what is observable. We shall now scrutinize the statistical and probabilistic paradigm and either confirm or refute their scientific nature.

1. Philosophical incompatibility with science:

Science is based on two believes, philosophical understandings, assumptions, principles that (first) the Universe is ordered, i.e. that it is both governed and driven by laws, which (second) are intelligible for us humans. Neither the order of the world, nor the intelligibility of its laws has to be so and neither of them can be proven by us. We cannot prove that the world is ordered or intelligible, as we cannot prove that whatever we see and understand from, and about it is indeed so. However we believe that the world is ordered and we also believe that it is intelligible. This belief is the foundation for science.

For example, physics and mathematics base themselves on this notion and as a result they produce laws about the Universe. Statistics and probability on the other hand do not study, understand and present intelligible laws with clearly specified type, semantics, boundaries, etc. in relation to the Universe. Whenever faced with any particular problem statistics and probability always starts with one or both of the following:

1. there is no law that governs the matter, thus denying the first of the two fundamentals principles of science that the world is ordered,

2. there is a law that governs the matter but we cannot formalize it, which is denial of the second fundamental principle of science that laws governing the world are intelligible,

then they proceed using some mechanical aggregation of some (whatever) available data, which may or may not be adequately related to the examined matter. Thus by its nature statistics and probability are unscientific because they deny one or both of the fundamental principles of science which again are that the world is ordered (law for everything) and intelligible (we understand all laws).

Therefore as the basis of science is the above philosophy and as statistics and probability clearly contradict these fundamental principles of science, it is necessary to conclude that statistics and probability are clearly NOT scientific subjects. They certainly use scientific notation which is necessary but by far not a sufficient condition for something to be science. Some particular examples will be presented later in the article.

It is indeed quite astonishing to note some of the more prosaic examples for the complete philosophical discrepancy of the subjects, i.e. the term "approximate truth" defined in statistics - truth by definition is either complete and only or it is not truth at all. Perhaps you can feel the blow of political correctness?!

2. Fundamental inconsistency:

2.1. Statistics and probability assume that an explicit law (for some matter) does not exist or at least cannot be formalized;

2.2. Although they deny the law of study, they use other laws (e.g. for existence and nature of numbers, operations for additions, subtraction, and so forth, including defining their own laws) - thus they imply that the world is not coherent, i.e. there are laws for some things and there are no laws for other things, however an incoherent world is not study-able at all as it is impossible to establish any foundations, boundaries and so forth;

2.3. Statistics (and to certain extent probability) presents its own conclusion as a true fact i.e. as a law.

Obviously:
- 2.3. contradicts 2.1 unconditionally;
- 2.2. could contradict 2.1 subject of circumstances;
- 2.2. alone implies that statistics and probability themselves cannot exist as a science because no science is possible in an incoherent world.

We are therefore required to conclude that statistics and probability are self-contradictory and void subjects as they try to study/describe a non-study-able by their own assumption Universe.

3. Erroneous and dangerous:

A field where statistics is heavily used is when "proving" the effectiveness of drugs. For example new drug is 90% effective in 123 cases, old drug is effective ... is the new drug better?

Lets us consider a broken TV set which is taken by the owner to a repair maker since its warranty has expired. The repair maker says that he does not have the scheme of this model, but for this problem he has noticed that if he add a new part to the TV set the device will start working in 90% of the cases. A TV set intact however works 100% of the cases, therefore the fix is somewhat related to the problem but is definitely not really fixing it. The owner however has no other choice and agrees to have the 90% fix put in place. Indeed after the "fix" the TV set operates again ... for a while because the result of the fix is not only that the broken part is still broken, but that some other parts of the system (TV set) were placed to function in conditions not optimal for them - that is because in a system intact all subsystems and elements are designed to work, and work in optimal conditions, and adding or removing anything destroys that optimal state. As a consequence after some short period of time the TV set breaks again simply because one or more overloaded elements gave up. Drawing the line, the system with originally relatively smaller damage became more damaged (possibly this time severely), simply because the electric laws were replaced with probabilistic and statistical nonsense and ignorance.

Is it then surprising that people are getting sicker and sicker after being mistreated with drugs always "proved" to be "working" statistically?! Because biological machines (such as humans) are much more complex than a TV set, with many more interrelated parts (at any level), buffers, closed loops, self-adjusting and self-repairing subsystems, these effects are slower to demonstrate themselves, however they do exist and do demonstrate themselves.


Consider a dice – a fair dice should account for a fairly flat distribution for all numbers when rolled - that is similar amounts of 1s, 2s, 3s, etc. Note that the word should is not correct as its presence there is not proven, but for now we will assume it. For example if I have 6 sided fair dice and I throw it 60 times I would expect to have about 10 ones, 10 twos, 10 threes, etc. occurred purely by chance. In reality these numbers will not be so perfectly distributed, e.g. any side may occur a few times more or less than the other sides. If one is to use a unfair dice some of the sides for which the dice is biased will occur somewhat more often than other sides. Statistics devices a technique known as "Fit Test" through which it claims it is able to prove if a dice is fair or unfair.

However I am able to quite easily create a dice with an embedded micro-controller, battery and mechanics which would perform as follow. From the very first roll of the dice the microprocessor and software will ensure that the dice sets to 1, on the second roll it sets to 2, on the third to 3, etc., to 6. On the seventh roll the micro-controller will set the dice again to 1 and so forth, thus this obviously unfair dice will have a perfect distribution. After each 6 rolls, the number of 1s, 2s, 3s, etc. will be all equal. According to the "Fit Test" a dice with such flat distribution is the "THE PERFECT FAIR DICE", but we know better that it is not, so "Fit Test" and the statistical paradigm has just once again failed.

A statistician will now object that the dice above will never roll on the same side in 2 consequent rolls, so he/she can "catch" it. There are two ways to respond:

- The first is that I can easily modify the program running on the micro-controller so that the micro-controller does not interfere on the first roll from any series of rolls, however it records the side occurred. On the second roll the micro-controller ensures that the side that was drawn on the first roll does not occur. On the third draw it ensures that none of the previously two draws occurs and so forth. Thus we could have the same side occurring 2 simultaneous times - the last roll from one group of rolls and the first roll from the next group. Of course this amendment of the software would not work for 3, 4 and more sequential same side rolls. However it is obvious that one is able to place whatever software he/she wishes that operates in a particular way so that it is not detectable by any "Fit Test". Examples are numerous, I will give just one – a random number generator from 1 to 6 running on the micro-controller. For a "Fit Test" this dice will be a an ordinary fair dice, but we know very well that random numbers in computers DO NOT EXIST, and every random generator although seemingly random is not random at all. Further, what if the random generator is re-seeded via a special radio signal? Or what if the micro-controller engages only in the presence of special radio signal, thus the proof that the dice is fake can only be made by cutting the dice and examining its internals, passing it through x-rays, or other such true scientific means.

- The other more general way to respond is to request a proof why should we expect to have 2 or more sequential draws on the same side of the dice? Based on the first fundamental assumption of science (that the world is ordered), I argue that if one is to roll the same dice in exactly the same way in exactly the same conditions he/she will always get exactly the same side. As no one accounts for the exact conditions in the so called roll "by chance", no one also have any (scientific, logical, or whatever valid) reasons to impose any particular expectations for the outcome, because none of the experiments (rolls) were actually properly observed.

This later argument obviously totally destroys all on what probability and statistics are based, and are about, and this is what science is about - proper and objective observation and experimentation, not some wishy-washy fairy tales. Once again we demonstrated that probability and statistics are simply non scientific disciplines.


Suppose two software programmers write two computer programs, both working independently and both attempting to solve the same problem. One of them does everything perfect but unfortunately makes a typo in the very last line of his program and mistakenly negates the result, thus always yielding in a wrong answer. The other programmer however makes a mistake in the algorithm, and it happens that his program gives correct results in 95% of the cases. A statistician is asked to make a "Fit Test" and select the better program. The statistician quickly presents his/her choice - the second - 95% of cases - correct program. The first program is then abandoned and possibly destroyed. However that is the wrong choice - a typo can be easily fixed, however a wrong algorithm might be impossible fix.

Conclusion

To prove something one must prove that it is always true. However to disprove it one need to demonstrate ONLY ONE instance where it is wrong. Even in an article as short as this one we clearly disproved any claim that statistics and probability are sciences. We showed this in three different ways by philosophy, by logic and by engineering using number of examples. Any one of these, on its own, and even any one from the engineering examples on its own is a sufficient condition to overthrow statistics and probability are sciences. Therefore we successfully confirmed the claim stated at the beginning that statistics and probability are most definitely not sciences, but are purely self serving methodologies yielding meaningless and unattached to objective reality results (for as much as science is objective based on its fundamental assumptions). The intricate problem of statistics and probability is that they both simply mangle some numbers, instead to understand the Universe of Discourse which is what the objective of science is.

Statistics and probability are the politically correct and politically misused pseudo scientific disciplines utilized to "prove" many things, including that DDT is safe, pesticides and herbicides are OK, drugs that work XX% will help you, and so on ... but do you remember the fix for the TV set above? I see two plausible reasons to explain the promotion of these unscientific subjects: 1. political and financial misuse, and 2. inability of some people engaged with science to discern the laws that they try to discover so they refer to probabilistic and statistical methods instead of working harder or leaving the fame for someone else, someone who has ideas.

An interesting suggestion is that statistician begins dropping apples until an apple does not fall, but instead it stays stationary in the air, and thus prove that the law of gravity is not a law and the statistical/probabilistic approach is more adequate.

Miroslav B. Bonchev
26-th August 2008