Absolute courses - page 33

 
Avals:

generate a course ED and DY using random walk and add EY=ED*DY.Then similarly find E,D,Y so that KK->1. What will they now display as a pattern of SB?

They won't. I guess my algorithm will collapse and will not be able to represent three similar curves with KK->1. The very possibility to reduce to a single shape is determined by the non-randomness of the quotes. I'll give it a try. Today, tomorrow, the day after tomorrow I will post it here. Actually, in my past incarnations here I myself have repeatedly suggested testing all sorts of algorithms on Gaussian white noise and on simple functions (sines, meanders, steps).
 

I don't know.... I don't see what your calculations would give compared to conventional indices. One has to go beyond some assumptions and approximations in the calculations. Regarding the JPY, the indices also show its decline from may 2012 to january 2013, the last few weeks this movement has slowed down, maybe even reversed. Same eggs on the other side.

 
Figar0:

I don't know.... I don't see what your calculations would give compared to conventional indices. One has to make some assumptions and approximations in the calculations. Regarding the JPY, the indices also show its decline from may 2012 to january 2013, the last few weeks this movement has slowed down, maybe even reversed. Same eggs on the other side.


How many times do I have to tell you, NOT THE SAME. What "indexes"? They're charts of the yen relative to what? Against a basket of currencies? The value of that basket of currencies changes wildly. You can't. You can't!!!!!!!!!!!!!!!!!!11 you can't!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!1 There has to be a REAL parrot. For example the yen itself in some bar on 1 January 2012. And so in relation to it, by definition equal to itself all this year with a ponytail and at all times, build.
 
Dr.F.:

Colleague, I gave you an example. Sine and cosine. Correlation is zero. No orthogonality. What else do you need?
Sine and cosine are functions of angle and are not random variables and are also rigidly related by identity (sin^2(a)+cos^2(a)=1), while correlation by definition is used as a measure of relationship for random variables. Therefore it makes no sense to talk about any correlation for these quantities.
 
Dr.F.:
They won't. I guess my algorithm will break down and will not be able to depict three similar curves with QC->1. The very possibility to reduce to a single shape is determined by the non-randomness of the quotes. I'll give it a try. Today, tomorrow, the day after tomorrow I will post it here. Actually, in my past incarnations here I myself have repeatedly suggested testing all sorts of algorithms on Gaussian white noise and on simple functions (sines, meanders, steps).



I don't believe it!!! I propose to do the following. Give you three sets of these pairs.

1- As Avals said to generate entirely from gpc

2- Unfamiliarize yourself with a chunk of history of 2 unknown pairs (similar to ED and DA, only with a different order of currencies in them) their normalized values.

3- Make gpsc with distributions from real pairs. from point 2.

4- One of the series is a real pair and the other is a gpsc.

 
khorosh:
Correlation, by definition, is used as a measure of correlation for random variables. Therefore, it does not make sense to speak of any correlation for these variables.


Another delusional person? Correlation (I mean C. Pearson's linear correlation coefficient, which characterises the existence of a linear relationship between two quantities) is used as a measure of the relationship of NONE, in general, by itself, VOLUNTEERS. Maybe random, maybe x and x squared, whatever. What do you care if they are random or not? You have a formula, substitute it, find the result. You say, "OK, the correlation coefficient between x and x squared in the interval x is almost 0.97. And what?
 
Joperniiteatr:



I don't believe it!!! I suggest you do the following. Give you a dash of a set of these pairs.

1- As Avals said to generate entirely from gpc

2- Unsuspect you a piece of history of 2 unknown pairs (similar to ED and DA, only with a different order of currencies in them) their normalized values.

3- Make a gpsc with distributions from real pairs from point 2.


Agree to the experiment :-)

Have someone prepare "real" EURUSD and EURJPY and "random". Let the "real" ones be just clause columns too, without everything else (dates, open, and other stuff). You can distort them somehow, but NOT CHANGE THE NATURE, e.g. take an exotic non-major triangle and invert the time direction. And I'll give you the results for both pairs of files.

 
Dr.F.:


Agree to the experiment :-)

Have someone prepare "real" EURUSD and EURJPY and "random". The "real" ones will also just be clause columns without everything else (dates, opener, etc.). You can distort them somehow, but NOT CHANGE THE NATURE, e.g. take an exotic non-major triangle and invert the time direction. And I'll indicate the results for both pairs of files.



But it will be real EURUSD and EURJPY not like 1.3333 and 125.00, it will be real rates in normalised form so you don't peek. Will this do?
 
Joperniiteatr:


But it will be real EURUSD and EURJPY not like 1.3333 and 125.00, it will be real rates in normalised form, so you don't peek. Will that work?
OK, fine. But I ask you AFTER I've answered your results to reveal the way the original data is distorted and to present the original data itself too. To make sure that your distortion algorithm does not distort the nature of quotes.
 
Dr.F.:

Another delusional one? Correlation (I mean C. Pearson's linear correlation coefficient, which characterizes the existence of a linear relationship between two quantities) is used as a measure of the relationship of NONE, ANY, EVERYTHING, EVERYWHERE, VOLUNTEERS. Maybe random, maybe x and x squared, whatever. Why do you care if they are random or not? You have a formula, substitute it, find the result. You say, "OK, the correlation coefficient between x and x squared in the interval x is almost 0.97. And what?

Read wikipedia:

Correlation(from Latincorrelatio), (correlation dependence) is a statistical relationship between two or morerandom variables(or between variables that can be considered as such with some degree of accuracy).