That in the "black box"?

The statistical approach

All of the above also applies to the version of the method <black box> that are focused on the use of statistical packages. However, these options also have specific features. The very possibility of the application of probabilistic methods require justification. Usually say: <This value changes do not understand how so with a clear conscience, we can assume its random>. And that's being built, all further arguments, using sometimes very subtle mathematical methods. But the premise is wrong. It should be a clear distinction between the behavior of the irregular and random variables. Random variables have some very special properties, and only to them can be applied probabilistic methods. In particular, if we calculate the average realizations of the random variable regularly arranged on a large set of values of time, must be obtained the same result, regardless of the set. For example, to a price of paper can be regarded as a random variable, it is necessary that the average price of all Tuesdays equal to the average price of all Thursdays. If we calculate the average price of all days, when the thirteenth falls on a Friday, then again, should have the same value.

Variable can be random if it has gone through many of these tests (in theory - infinitely many). If at least one such test has not converged - the value is not random. We are not aware of any case that analysts using probabilistic methods in modeling financial market, conducted a similar test.

Here he writes about the theory of probability, Bertrand Russell [1]: "Mathematical possibility always arises from a combination of two statements, one of which can be fully known, and the other is completely unknown. If I take the card out of the deck, what is the chance that it will be an ace? I totally know the structure of a deck of cards, and I know that one out of every thirteen cards have an ace, but I do not know what card I'll get>. In order parameter, which is completely unknown, it is usually all right. But with others - often the situation is much worse. But for this it is not accepted to pay attention.

Justifying the use of probabilistic methods, most often nod to their successful use in the natural sciences. But there's initial hypotheses or carefully checked experimentally (as in the theory of shooting), or theoretically justified (as in statistical physics). Somehow, in the economy or any other is not considered mandatory. In order to use probabilistic methods were justified, as required, hours to the data series were long enough. Many authors note that, in this case their data is not enough ... and yet are probabilistic methods.

Do not read <Short Course ...>

Even more obvious mistakes are made when it comes to choosing the law of the random variable. To quote [2]: <Interference these [the information flows] are the result of a large number of random factors, and this makes it easier to work with them, If the distribution is known - is normal, according to the central limit theorem of probability theory>.