Bayesian regression - Has anyone made an EA using this algorithm? - page 32
You are missing trading opportunities:
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
Registration
Log in
You agree to website policy and terms of use
If you do not have an account, please register
To write a program, please advise whether the normal distribution of results is available in MT4, or whether to use another one.
I am amazed at the high level of mastery of mathematical methods by the panelists and their complete lack of understanding of the principles of their applicability. Any regression analyses correlated data. If there is no correlation, then the regression is not applicable. If the distribution of the quantities under study is different from normal, parametric statistics methods are also not applicable. The market does not have the property of normality. Also the market as a process doesn't depend on time. Both of these things, however, defeat the very idea of regression analysis at its root.
I am amazed at the high level of proficiency in mathematical methods of discussion participants against a background of complete ignorance of the principles of their applicability. Any regression analyses correlated data. If there is no correlation, then the regression is not applicable. If the distribution of the quantities under study is different from normal, parametric statistics methods are also not applicable. The market does not have the property of normality. Also the market as a process doesn't depend on time. Both of these things, however, defeat the very idea of regression analysis at its root.
Well, finally, the voice of reason.
In the old days, the study of applied mathematics began with the study of systematic errors of the first and second kind, the meaning of which was taken from systems analysis rather than from statistics.
The first kind of systematic error was formulated as follows:
The correct application of correct methods to data to which these methods do not apply.
The basis of application of mathematical methods in general and of statistical methods in particular is the REASON of applicability of these very methods. And nowadays the importance of this very justification has increased repeatedly in connection with wide access to the most sophisticated mathematical tools in the form of software packages: it is not necessary to understand the internal construction of the method - a couple of lines and all. But to justify the application of....
The normal PRNG generates uniformly distributed numbers. To convert a uniform distribution to a normal distribution, a special conversion algorithm must be used.
I am amazed at the high level of mastery of mathematical methods by the panelists and their complete lack of understanding of the principles of their applicability. Any regression analyses correlated data. If there is no correlation, the regression is not applicable. If the distribution of the quantities under study is different from normal, parametric statistics methods are also not applicable. The market does not have the property of normality. Also the market as a process doesn't depend on time. Both of these things, cross out the very idea of regression analysis, whatever it may be at the root.
And I'm surprised at the inconsistency in the posts of participants with a high level of competence. Recently, in another thread by Illita, you confirmed the existence of a normal distribution. True, there it was about the spread and you wrote: "Distribution analysis is of interest only from the point of view of studying trading conditions. There is no fish here". And now you write "The market doesn't have the property of normality".
I'm not the one who wrote about the fact that volatility, increments, have a distribution law close to normal and gave charts. I just took it into consideration as I believe in it.
In general, I am interested in the Bayesian approach itself and the attempt to calculate a probability measure as a product of probabilities using Bayes formula. And it's up to everyone to build a regression on it. I think there are fish in the water here.
Well, finally, the voice of reason.
In the old days, the study of applied mathematics began with the study of systematic errors of the first and second kind, the meaning of which was taken from systems analysis rather than from statistics.
The first kind of systematic error was formulated as follows:
The correct application of correct methods to data to which these methods do not apply.
The basis of application of mathematical methods in general and of statistical methods in particular is the REASON of applicability of these very methods. And nowadays the importance of this very justification has increased repeatedly in connection with wide access to the most sophisticated mathematical tools in the form of software packages: it is not necessary to understand the internal construction of the method - a couple of lines and all. But to RECOGNISE the application of....
Thanks for the reminder of systemic errors. In your brief history of technical analysis in a post on this thread you wrote: "The place of Bayesian models in financial markets is long and clearly defined - not applicable".
It is very interesting how Bayesian models were applied and who determined immutability. Bayesian methods are widely used in fraud detection, spam, medicine. Why do you reject them in forex?
I want to quote from a Habra discussion on Bayes.
"It is probably worth saying that such methods, when designing algorithms, require a fairly high mathematical culture of the developer, since the slightest error in the output and/or implementation of computational formulas will nullify and discredit the entire method. Probabilistic methods are particularly prone to this, because human thinking is not adapted to work with probabilistic categories, and therefore there is no "visibility" and understanding of the "physical meaning" of intermediate and final probabilistic parameters. Such understanding exists only for the basic concepts of probability theory, and then you need only very carefully combine and derive complex things according to the laws of probability theory - common sense for composite objects is no longer helpful. In particular, quite serious methodological battles that take place in modern books on philosophy of probability, as well as a large number of sophisms, paradoxes and puzzles on this topic are related to this".
Thanks for the reminder of the systemic errors. In your brief history of technical analysis in a post on this thread you wrote: "The place of Bayesian models in financial markets is long and clearly defined - not applicable."
Very interesting how Bayesian models were applied and who determined immutability.
Read my post again.
Same thing, but in different words.
Each matmethod is applicable to very specific data, so the applicability of Bayesian is not determined by anyone, but by the data to which it is applied. Several posts have been devoted to this issue.
And to put it even more simply, a screwdriver to screws and a spanner to bolts.
Once again read my post. The same, but in other words.
Each math method is applicable to very specific data, so the applicability of Bayes is not determined by someone, but by the data to which it is applied. Several posts have been devoted to this issue. And to put it even more simply, a screwdriver to screws, and a spanner to bolts.
Reread your post againhttps://www.mql5.com/ru/forum/72329/page17 I find it hard to argue. Let me ask a question.
It has been shown here that price increments have a distribution law close to normal. Do you disagree with that?
I want to use this as an a priori probability in the Bayes formula. Is that wrong?
P.S. " And even more simply, a screwdriver to screws and a spanner to bolts." Modern good screws have hex head wrenching (when it's hard to work with a screwdriver) and good bolts have screwdriver slots (when you can't reach with a wrench). Please understand this both literally and figuratively. My point is that this data (screws and bolts) is very diverse in nature. I don't think that the data from the "battle" floors of exchanges (with which the classical technical analysis operates) is adequate for Forex. In Forex, unfortunately, there is a game simulation of the real market.
I want to use this as an a priori probability in the Bayes formula. Is that wrong?
I don't do Bayesian regression.
I'm a professional mathematician, maybe a bad one, BUT to me, the usual steps are for me for any models:
Or we can simply keep in mind (as written above) that application of regressions in the market is an intriguing thing, God willing, that the price increments, not the prices themselves, will fit into some GARCH.
I am amazed at the high level of mastery of mathematical methods by the panelists and their complete lack of understanding of the principles of their applicability. Any regression analyses correlated data. If there is no correlation, the regression is not applicable. If the distribution of the quantities under study is different from normal, parametric statistics methods are also not applicable. The market does not have the property of normality. Also the market as a process doesn't depend on time. Both of them crosses out the very idea of the regression analysis, no matter what it is at the root.
Regression analysis does not require a normal distribution of the input data, it requires a normal distribution of the model residuals.
All economic data, price characteristics, etc. are correlated. There is no uncorrelated data.
Price is time dependent.