Discussion of article "Probability theory and mathematical statistics with examples (part I): Fundamentals and elementary theory" - page 7
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
The schedule's different, I'm just sorting out the provided hardware.
That's why the question arose, what do we get if my values are my values according to Bernouli's scheme?
Try to check the resulting division into two parts by means of the script from the second example about hypothesis testing (Fisher's exact test). It tests whether the difference between p1 and p2 is significant.
Try to check the resulting split into two parts using the script from the second hypothesis testing example (Fisher's exact test). It tests whether the difference between p1 and p2 is significant.
I'll be up late tomorrow night to look into it further.
my question is a beaten question... I have repeatedly tested articles on understandable data (Weierstrass function), as practice shows, very often authors of articles use mathematical apparatus from different fields, but exclusively on EURUSD ..... in general, if it takes a long time to explain, we will come to the article on Hubre, which I mentioned https://www.mql5.com/ru/forum/345555/page4#comment_17178481.
script to generate my symbol attached
ZY: nothing personal, but looking for the truth! )))
The schedule's different, I'm just sorting out the provided hardware.
That's why the question arose, what do we get if my values are my values according to Bernouli's scheme?
for this graph:
n1e=34069 p1e=0.5006604244327688 p2e=0.09090909090909091
D1 chart itself, where the trend is clearly present
I think you have some kind of sine wave there.)
The model from the fourth example clearly does not fit here because of the obvious dependence between the increments. This is also clearly seen by the fact that it "cut off" a very small tip.
Here it is better to try the model from the fifth example (Markov chain). You just need to choose the price sampling step such that the difference between p1e and p2e is maximal.
PS The model from the fourth example + Fisher's test can also be applied here. Only it should be applied not to the whole piece of price at once, but for example, at each appearance of a new bar. When the split into two pieces is confirmed, the older piece is discarded from consideration, etc. This is a simple version of what is called "online discrepancy search"
ZY: nothing personal, but looking for the truth! )))
I'm all for it))
You've got some kind of sine wave going on there, I think.)
not a sine wave, run the script, it will do it all by itself
For the Markov chain model (5th example, script "markov_model.mq5") it is necessary to take a price sampling step of 5%, then the probability estimates are maximally distant from each other and equal to about 0.4 and 0.6.
It turns out (as it should be) antipersistence - a change of direction is more probable.
I am merely disagreeing with your original assertion that piecewise constant models are inapplicable, and arguing that (from a very general point of view) they are the only ones we use.
In theory, yes. In practice, if the frequency of model recalculation is much higher than the frequency of trades, we can consider that piecewise constant does not affect the TS.
Alexey, can you elaborate on the maximum likelihood method?
Suppose we are building a linear approximation. If I remember correctly, when the data are normally distributed, the best approximation is the MNC, and when the data are tailed, the best approximation is the MNM. And both derive their conclusion from the MMP. Is it possible to explain all this in a simple way?
Alexei, can you elaborate on the maximum likelihood method?
Suppose we are building a linear approximation. If I remember correctly, when the data are normally distributed, the best approximation is the MNC, and when the data are tailed, the best approximation is the MNM. And both derive their conclusion from the MMP. Is it possible to explain all this in a simple way?
Everything is correct, but not just "tailed", but for example Laplace. Maybe some others, but certainly not Cauchy's, for example, which is much more "tailed".
In the next article I'll deal with something like this - up to a numerical example. In principle, there is a usual problem on extrema of a function (solved by searching for the zero of derivatives).
I used to read the archive of his posts on cyber spider with interest. Unfortunately, not everyone can possess such a sharp mind as the deceased had. Personally, I prefer Alexander Gorchakov's boring but understandable lectures.
On the topic of the quotation at your link - in my opinion it is an attempt to say in probabilistic language (Prigogine, Bayes formula, etc.) those things that nowadays econophysics tries to say in the languages of game theory and statistical physics - phase states and their changes, etc., etc. Moreover, econophysics says this exactly about financial markets, without resorting to strains in the form of analogies with biological objects.
A ~6th generation mathematician, where is he to him.
On a household level (as I understand it), he writes about cyclicality/clustering of volatility and the importance of senior tf.
Interesting that he tried to calculate this mathematically, not just an abstract description.
That's what prevents them (Ataman, Ilyinsky) from making an introduction, in simple language, so that you don't have to reread each line 10 times.