Dependency statistics in quotes (information theory, correlation and other feature selection methods) - page 27
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
Это - ряд остатков взаимной информации реального и случайного ряда:
Frankly, it's not much of a stretch that the residual is incredibly similar in structure to the effects of volatility.
The idea is good, here is a generated SB chart with identical volatility to EURUSD. Alexey, would you be so kind as to perform an analysis for it. Let's see if there are any differences.
You should not take a single volatility, but for EURUSD 3 or so (it depends on the DC filters). Because ticks are not only +1/-1 points, and SB is generated by one point per tick.
Frankly, it is not much of a stretch that the residual is incredibly similar in structure to volatility effects.
If volatility is accounted for correctly, then it is memory effects for the sign of price change on the zero bar. If there are problems with volatility, then it is residual volatility + other dependencies.
The question is: how do we make sure we have accounted for all the volatility?
If volatility is accounted for correctly, then it is memory effects for the sign of price change on the zero bar. If there are problems with volatility, then it is residual volatility + other dependencies.
The question is - how do we make sure we have accounted for all the volatility?
О! I have an idea. I'll try to take the initial EURUSD H1 series today, take its returns and just randomly mix up the signs of price change. Eureka, that will be the exact volatility with a coin-like up/down price move. What remains then, when comparing the data, is the net sign dependencies.
Am I thinking correctly? What do you suggest?
Frankly, it is not much of a stretch that the residual is incredibly similar in structure to volatility effects.
you don't have to take a single ox, but for EURUSD 3 or so (depends on filters of the brokerage company). Because ticks are not only +1/-1 points, and SB is generated with one point per tick.
Taking into account tick effects could be a second step, say laying down real returns.
I wonder if Yurixx is still looking in here, I remember he was just doing this sort of thing
О! Here's an idea. I'm going to try to take the initial EURUSD H1 series today, take returns on it and just randomly mix up the price movement signs. Eureka, that will be the exact volatility with a coin-like up/down price movement. What remains then, when comparing the data, is the net sign dependencies.
Am I thinking correctly? What do you suggest?
I think that not every volatility is harmful for us, and ideally we may benefit from the direction and magnitude of price movement. So it's hardly worth fighting to eliminate volatility altogether, imho. I suggested another approach (normalization to the profile), but since you have chosen it, first of all, we should check several realizations before drawing conclusions, and secondly, we should not dwell on volatility, it is not the only known effect, the same recoverability is also an established property of price series.
Normalising to the profile still leaves correlations, I've tried it. Also, there is a strong dependence on volatility at the nearest lags, and normalisation does not completely eliminate this.
And the return is what (as far as a number of returns are concerned)? If there was a + will follow - with a slightly higher probability?
О! Here's an idea. I'll try to take the initial EURUSD H1 series today, take returns on it and just randomly mix up the price movement signs. Eureka, that will be the exact volatility with a coin-like up/down price movement. What remains then, when comparing the data, is the net sign dependencies.
Am I thinking correctly? What do you suggest?
yes, also an option :) you can even choose a sign at random - "+" with a probability of 0.5 and "-" with the same probability
волу надо не единичную брать, а для EURUSD 3 где-то.
I seem to have generated by default, i.e. the coefficient was 8. What also surprised me was that the SB amplitude was several times greater than the real one.
I suggest to deal with one sample of EURUSD_RAND for the time being, and when the issue is resolved, connect the other samples as well.
Normalising to the profile still leaves correlations, I've tried it. Also, there is a strong dependence on volatility at the closest lags, and normalisation does not completely eliminate this.
And the return is what (as far as a number of returns are concerned)? If there was a + will follow - with a slightly higher probability?
For ticks the minus would follow with a nа much more likely