A formal definition of a master-slave - is there one? - page 3

New comment
 
Cmu4:

There's talk about correlation... what method do you use to measure it?

There are many, not all of them are suitable.

Correlation is not suitable at all. It is for stationary rows, and there are no stationary rows in forex
 

Why is it for stationary?! It is actually a quantity that characterises the relationship between two random variables.

From this, of course, it can be deduced that its value can have significance = 0 at all.

 
Cmu4:

There's been talk about correlation... what method do you use to measure it?

There are many, not all of them are suitable.


I measure it with Spearman.
 
faa1947:
Correlation is not suitable at all. It is for stationary rows, and there is no such thing on fore.

You have made a mistake. Correlation is suitable for any series. It is Fourier and regression only suitable for stationary series.
 
Cmu4:

There's been talk about correlation... what method do you use to measure it?

There are many, not all of them are suitable.

What's the right method? Is Pearson appropriate? The general formula without expectation and variance estimates seems very logical.
 
wmlab:

You've got something wrong. Correlation is suitable for any series. It is Fourier and regression only suitable for stationary.
I think not. Cointegration is more general and that with limitations on application. I don't want to look at it. I'm just sure correlation is not applicable at all on fore. It's a number. What place in the sample does it refer to? And we are generally interested in the right-hand edge of the sample.
 
faa1947:
I think not. Co-integration is more general and has limitations on its application. I don't want to look at it. I'm just sure that correlation is not applicable at all on fore. It's a number. What place in the sample does it refer to? And we're generally interested in the right-hand edge of the sample.
The one you specify in the series you're comparing.
 
GaryKa:
What methods are suitable? Is Pearson suitable? The general formula without expectation and variance estimates seems very logical.
Pearson is unlikely. How you calculate it depends on what you want to get.
 
Cmu4:
To the one you specify in the series you are comparing.

Correlation has no place in a series - it is a characteristic of a sample of two series.

Correlation is the biggest illusion in statistics for people who not only know statistics, but feel it.

If we talk about forex, we cannot simply apply it, because forex has trends and correlation values indicate the ratio of two deterministic components in two series, i.e. have nothing to do with random variables. So excuse me, all arguments about Pearson's and Spearman's here are from the evil one.

 
Cmu4:
Pearson's is unlikely. How you calculate it depends on what you want to get.

See, if I understand you correctly, then Pearson is "unlikely" to fit, because it is used to estimate a measure of linear relationships, and hence is not suitable for estimating a measure of non-linear relationships.

But in that case you can:

  1. Either "non-linear" transform the input data (another question is how and why exactly) before applying it to Pearson's input
  2. Or to introduce "non-linearity" into the formula itself (where the scalar product is), but it will be a "slightly different" Pearson coefficient))

The idea of using the normalized expectation of this relationship as a measure of the stability of the relationship is acceptable to me.

12345678910
New comment
Reason: