Dependency statistics in quotes (information theory, correlation and other feature selection methods) - page 17
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
As a method of elimination, I suggest simply relating the increments to the daily volatility profile.
Can I put it in my own way?
So, the chosen approach shows that there are dependencies. The most obvious, reasonable and visible to the naked eye is the daily periodicity of volatility.
Therefore, the next logical step in my research would be to try to exclude this obvious and very strong dependence from the data and see if our (your) method shows the presence of other dependencies.
As a method of elimination I propose to simply relate the increments to the daily volatility profile.
I'm terribly sorry.
but what has "WRONG" volatility got to do with it if we've adopted the "correct" model.
:)
I continue to "destructively" argue that the test for independence is equivalent to the test for a uniform distribution.
And no "non-parametric statistics" - just the null hypothesis, which textbook authors are sometimes too lazy to explain...
Rough code for MQL-- https://www.mql5.com/ru/forum/132692/page13
Dear Expert!
There is a question about this machine. As a result of devolatilisation, is the autocorrelation on the resulting series of returns (taken modulo) close to zero? In a normal process on lag 1 and 24 the autocorrelation is around 0.11.
I can certainly check it myself, it's just that I tried to do a correction to the daily volatility profile myself, but the autocorrelation for some reason remained.... And this is the root of the strong dependencies, as has become clear.
Dear Expert!
Don't call me that :) TheXpert is a nickname, nothing more, expert is a characteristic.
As a result of devolatilisation, the autocorrelation on the resulting series of returns (taken modulo) are close to zero?
I have no idea, for me the daily smoothed ATR is a purely practical tool, and it didn't go further than getting a graph, more pressing matters came up.
So you have to do it :). Not necessarily close, but logically they should be.
I'm terribly sorry.
But what has "WRONG" volatility got to do with it if we've adopted the "right" model.
:)
Did I just walk over the moon or did you? :) Why is your volatility wrong and indeed, what does your wrong volatility have to do with it? It is your right to accept the model and consider it correct, but in relation to the author's approach any model would be external, there are no models in his approach and there cannot be. If I understand correctly of course :)
Did I just walk over the moon or did you? :) Why is your volatility wrong and, indeed, what has your wrong volatility got to do with it? It is your right to accept the model and consider it correct, but in relation to the author's approach any model would be external, there are no models in his approach and there cannot be. If I understand correctly of course :)
May I ask?
Do we understand independence in the same way? i.e. both processes belong to the same distribution and are supposedly independent.
But what if they are not the same?
what then?
hence "irregularity" .
:)
may I ask?
do we understand independence in the same way? i.e. both processes belong to the same distribution and are supposedly independent.
what if they are not the same?
what then?
hence the "irregularity" .
:)
I don't have time to acclimatise so quickly :). What two processes? There could be a million processes, their distributions could be whatever, we only see the overall result.
Volatility and its daily periodicity is just an observable fact, which has nothing to do with any model at all. Therefore it is always correct :).
I don't have time to acclimatise so quickly :). What two processes? There can be a million processes, their distributions can be whatever they want, we only see the overall result.
Volatility and its daily periodicity is just an observable fact, which has nothing to do with any model at all. Therefore it is always correct :).
you have returns (and Alexei claims that they are almost Laplacian in time distribution).
Now you test hypotheses about their independence from previous values.
If the pattern of distribution of returns is uniform - it is correct to apply chi-squared as discussed here...
Otherwise it's not. That's what I'm talking about. You have to take the frequency by the Laplace distribution for the chi-squared test. And think nothing else.
And the fact that volatility is sensitive to equity-volume is a fact. but the reason is different.
And trying to normalise - would blur the obvious cut.
The further out (over sigma) - the more independent...
;)
you have returns (and Alexei claims that they are almost Laplacian in time distribution).
Now you test hypotheses about their independence from previous values.
If the pattern of distribution of returns is uniform, correct. Otherwise it isn't. that's what I'm talking about.
And the fact that volatility is sensitive to equity-volatility is a fact. but the reason is different.
And trying to ration it - would blur the obvious cut.
The further out (over sigma) - the more independence...
;)
SVs can be distributed differently and can be dependent or independent. If 2 SVs are independent, then the conditional distributions of the independent random variables are equal to their unconditional distributions. In the case of one SV, the distribution is independent of previous SV values: the conditional SV distribution (of previous values of the same SV) is equal to the unconditional SV distribution