Random Flow Theory and FOREX - page 9

New comment
 
There's definitely a division of 0 in the log. Thank you.
 

Prival, I looked through your document, now I understand why you need linear regression. The good news is that what is described there is called a channel, about such channels people were engaged in the famous in narrow circles thread on a parallel forum :). I don't know about others, but I was just happy to use the Y - mu value :). IMHO, your idea in those terms can be reformulated as follows: use ACF to evaluate channel stability (i.e. to select and follow them). Like apparently most other participants, I am extremely sentimental about everything related to that topic :). That is, if the computational cost is acceptable, it would be interesting.

An extraneous question: The impression is that your indicator gives a predominantly positive phase DFT . Does this make sense physically?

 
lna01:

Prival, I looked through your document, now I understand why you need linear regression. The good news is that what is described there is called a channel, about such channels people were engaged in the famous in narrow circles thread on a parallel forum :). I don't know about others, but I was just happy to use the Y - mu value :). IMHO, your idea in those terms can be reformulated as follows: use ACF to evaluate channel stability (i.e. to select and follow them). Like apparently most other participants, I am extremely sentimental about everything related to that topic :). That is, if the computational costs are acceptable, it would be interesting.

An extraneous question: The impression is that your indicator gives a predominantly positive phase DFT . Does this make sense physically?

Unfortunately, I haven't read and don't know the parallel forum thread. I don't have time to do and read everything, moreover on the boundless expanses of the net. Thanks that at least someone writes, it seems to me all the time that 2-4 people are interested in this thread. I guess I understand why many are not interested, do not understand where we're going ;-(. I briefly described the way I did when I did a selection of books for a mathematician in stochastic resonance branch.

I'll try to elaborate on what all these exercises and our efforts are for. Just have to run off to work at the moment. What we have got has nothing to do with the channel (if I understand the channel correctly). About the phase, clarify the question.

I will try to make my thoughts based on this principle.

The goal without a way to achieve it is a mirage, the path without a goal is a road to nowhere.

 
Prival:
What you get has nothing to do with the channel (if I understand the channel correctly). Clarify the question about the phase.

That depends on what you call a channel :). IMHO, the constancy of the RMS is just a characteristic of a channel. However, I won't insist on the term or on the fact that I have understood the idea completely.

As for the phase: take the Pvr42 indicator, set h = false, watch and see that the result is mainly positive. At least for some reason I have come across precisely such graphs.

 

About the phase.

As always, I'll try to use an example. I made a quick optimal filter (file attached). By the way, I recommend it as a visual aid for Fourier Transform filtering. Pull the strings and see what happens.

Explanations of the program.

  1. A signal Y=5*cos(2*pi*10*t+0), a monochromatic signal with amplitude 5V, frequency 10Hz and phase 0, is modeled.
  2. The Fourier transform is performed and the optimum filter is derived from it.
  3. Why it is optimal. Note on the filter output Amplitude is 5V, experiment, give any amplitude at the input, the output will exactly equal let's say 345V. This happens because all the energy of the signal is collected in this filter.
  4. Experiment with the frequency of the signal, if you set the frequencies to an integer number and the Kotelnikov theorem holds, then everything is OK. The energy will again be collected in the same filter, as the filter is tuned precisely to this frequency. But if you set the frequency to a fractional number (say 10.76), i.e. there is a mismatch with the filter tuning frequency, the energy is distributed to all filters through the side lobes. The AFC of any filter is sin(x)/x. Construct this function, and you'll see what I'm talking about.
  5. Regarding the phase, although in the input signal phase =0, at the output of the 10th filter it is 1*e-13. This is due to the accuracy of the calculation (low-order error), it just accumulates as a result of conversions.

To Candid Phase in general is a very interesting thing, it's the only substance I know, that has speed of propagation higher than speed of light (it's just digression, if you interested can find mathematical proof of this phenomenon). Concerning positive values, most likely it is caused by two factors. The first is the features of the DFT try sin or cos. One will give + the other -. Second, where does such a huge number of signal components with the same phase (phase direction) come from. Try in my example to feed Y=t, i.e. a straight line equation, into the input. I think you will see, the subtraction of the 0th component of the spectrum in the indicator is not done quite correctly (I think you should Y-mu again :-) ). My attempts to use different Hemming and Butterworth windows didn't lead to anything, only made it worse (my teacher was right, application of windows is a tractor for energy). Therefore I've left the indicator in its present state. I can't remember, but I've told somewhere that this 0 component is still going to cause us some bloodshed :-).

All naturally IHMO, need to check and experiment. In general, according to various estimates, not taking phase into account in processing is a loss of about 30% of the information in the signal.

Files:
opt_filtr.zip  44 kb
 
Prival , Vinin has already written to you that usually the phase here is understood as something other than the initial phase of the harmonic oscillation. The phase in a process represented by a price chart in time is rather a stage (time interval) having certain properties. For example, if the linear regression coefficient is close to zero in the chosen period (time interval), we often speak about the consolidation phase (stage), calling it a flat. And when the coefficient is greater than a certain threshold value - another phase, calling it a trend.
 

Has anyone heard of the Dickey-Fuller (DF) stationarity test? It's all about my sheep... If you've heard of it, please let me know what you think.

2 Prival:

Phase in general is a very interesting thing, the only substance that I know, has a speed of propagation exceeding the speed of light

In STR it kind of proves that it is of no use, as it is not a signal - and moreover not a substance.

 
Mathemat:
Has anyone heard of Dickey-Fuller stationarity test? It's all about my sheep. If you've heard of it, I'd like to know your opinion.


There was something like that:

Dickey-Fuller criterion

Dickey-Fuller criterion is actually a group of criteria united by one idea and proposed and studied in [Dickey (1976)], [Fuller (1976)], [Dickey, Fuller (1979)], [Dickey, Fuller (1981)]. The hypothesis to be tested (null) in the Dickey-Fuller test is the hypothesis that the studied series xt belongs to the DS class (DS-hypothesis); an alternative hypothesis is that the studied series belongs to the TS class (TS-hypothesis). The Dickey-Fuller test actually suggests that the observed series is described by a first order autoregressive model (possibly corrected for a linear trend). The critical values depend on which statistical model is estimated and which probability model actually generates the observed values, with the following three model pairs (SM, statistical model; DGP, data generating process) being considered.

This or that?

 
It looks like it. Here's what I found: http://hometask.boom.ru/economics/econometrica/5.html .
 
rsi:
Prival , Vinin has already written to you

Thank you I have seen, only I have slightly different approach to the analysis. If you do not bother to read this branch. Unfortunately I have to discard all the concepts that do not have a rigorous mathematical description. Since it leads away, this iron (the computer) is not possible to stick "philosophical" concepts (trend, flat), etc. My opinion there are none !!!

To Candid, I hope I gave the right answer to him. I think I got the gist of the question right.

Thank you for your interest in this topic.

12345678910111213141516...85
New comment
Reason: