Building a trading system using digital low-pass filters - page 6

New comment
 
I propose to put forward ideas on methods of transition of rows to stats. I can check them and put into EA the filters obtained in this or that research. The final goal of cooperation is to get the algorithm of correct transition.
And all the rest (filters + EA) is already a matter of technique. In principle I don't mind to help.
 
there are at least 5 options
 
Integer:
there are at least 5 options

Let's consider everything step by step and check it out. I just see in this idea a great sense as even just spectroanalysis of nonstationary series in areas close (far conditionally) to the stationary one gives results.... I backtest and monitor 2 systems in real life
 
I kind of didn't say I was. I'm not trying to get information out of who you are. I myself have 2 degrees + post-graduate studies + work on eliptic integrals in serious scientific journals + work in the field of magnetism etc. I did not defend my dissertation only because I did not see fit to waste my time and went to Canada. It is simply easier to work as part of a team as you always need an outside view.
 
mql4-coding писал (а): I propose to put forward ideas on methods for moving from series to stationary.
The main question is the criterion of the stationarity of the series. Not a homebrew one, but a well-grounded one. As many times as I've tried it on forums, including mechmaticians, I haven't heard the answer. I got an impression that such a criterion, which is generally accepted, allegedly does not exist. I am ready to listen to variants as I do not believe that there is no such a criterion. The topic is too hot.

P.S. To begin, for example, with the simplest question: how it is possible to establish stationarity in the weak sense for some series, which is, say, a statistical constant with m.o. = 0 (sorry for poor terminology, as I am not an expert on statistics)?

It is obvious (to me) that m.o. alone is not enough to establish its stationarity in the weak sense. It is also necessary to know its s.r.o. It means that if the s.r.o. of the real series is equal to 0.5, the criterion which is based on the s.r.o. must not exceed 0.03, will reject stationarity of the studied series with a certain high probability, right? mql4-coding, help, eh?
 
Mathemat:
mql4-coding wrote (a): I propose to put forward ideas on methods for transitioning rows to stats.

The main question is the criterion of stationarity of the series. The main question is the criterion of stationarity of a series. How many times I tried it on the forums, including mechmatyans, - the answer was not heard. I got an impression that such a criterion, which is generally accepted, allegedly does not exist. I am ready to listen to variants as I do not believe that there is no such a criterion. The topic is too hot.



P.S. To begin, for example, with the simplest question: how it is possible to establish stationarity in the weak sense for some series, which is, say, a statistical constant with m.o. = 0 (sorry for poor terminology, as I am not an expert on statistics)?



It is obvious (to me) that m.o. alone is not enough to establish its stationarity in the weak sense. It is also necessary to know its s.r.o. It means that if the s.r.o. of the real series is equal to 0.5, the criterion which is based on the s.r.o. must not exceed 0.03, will reject stationarity of the studied series with a certain high probability, right? mql4-coding, help, eh?

In my opinion, the decreasing sequence of the mean (sko) is (may be) an estimate of the predictability of the series. The decreasing sequence of the variance of the mean. That is, the sko must be decreasing. For example if we consider a linear regression channel then the sko will decrease in it. I can also think and somehow insert Hearst criterion into series estimation (I'm not sure). I honestly didn't start out from definition of stationarity of series but considered it non-stationary at first :-) and was searching for correct ways of its analysis. But perhaps you are right and I should begin
I will think about it tomorrow and formulate stationarity criteria with formulas.
 
P.S. To begin with, for instance, a simple question: how can one establish stationarity in the weak sense for some series which is, say, a statistical constant with m.o. = 0 (sorry for the poor terminology, as I am not a statistician)? <br/ translate="no">

I don't understand what the problem is actually? There is a clear mathematical concept of a stationary random process - it is a random process whose probabilistic characteristics do not change over time.

Since in this case we are considering a time series, which is the realization of some random process, the notion of a stationary time series is quite obvious from the definition of a stationary random process.

Hence the conclusion that you're a bit mistaken. m.o. may not be equal to 0, and s.k.o. may be whatever. As long as these quantities are constant, the random process is stationary.
 
bstone:
P.S. To begin with, for instance, the simplest question: how can one
establish stationarity in the weak sense for some series,
which is, say, a statistical constant with m.o. = 0 (sorry for the
poor terminology, as I am not a statistician)?



I don't understand what the actual problem is? There is a clear mathematical concept of a stationary random process - it is a random process whose probability characteristics do not change over time.



Since in this case we are considering a time series, which is the realization of some random process, the notion of a stationary time series is quite obvious from the definition of a stationary random process.



Hence the conclusion that you're a bit mistaken. m.o. may not be equal to 0, and s.k.o. may be whatever. As long as these quantities are constant, the random process is stationary.
m.o. Is this the mathematical expectation?
 
m.o Is that a mate waiting?

yes
 

A random process (SP) with finite variance is called stationary in the broad sense if, its OLS (m.o.) and covariance function are invariant with respect to time shift, i.e. the OLS is constant (not time dependent) and the covariance function depends only on the difference of arguments t 2- t 1.


In some cases (which seems to me to be our forex case) a non-stationary process can be transformed into a stationary one.


Obviously it reduces to stationary. Most probably we are dealing with so-called periodically stationary or cyclostationary process.


Mathemat I gave you Tikhonov, it seems to have it all

12345678910111213...32
New comment