How do you calculate correlation in complex numbers? - MQL4 and MetaTrader 4

Vladimir Gomonov 2010.10.04 22:19 #111

Folks, how do you calculate correlation in complex numbers? If I stupidly transfer all Pearson's correlation calculations (for example) to the complex area, will anything meaningful come out at all?

I was looking on the Internet on this subject. In general, the calculation of correlation for complex variables is quite applicable (I found on the web mainly examples of butts in radiolocation and other radio communications).

I've been circling around this topic for a long time.

Are there any DSP experts here?

Study1: multi-currency analysis for Bayesian regression - Has Can I view the

hrenfx 2010.10.04 22:25 #112

Correlation analysis in complex variable econometrics

Vladimir Gomonov 2010.10.04 22:28 #113

hrenfx:

Correlation analysis in complex variable econometrics

Thanks, I'll go and have a look.

Dmitry Fedoseev 2010.10.04 22:56 #114

hrenfx:

For example, you 1. have a correlation coefficient calculation formula which is not obvious to me 2. ( corr2 function below).

...

2. If something is not obvious to you, that is your personal problem.

1. Look here, and apologise.

hrenfx 2010.10.04 23:15 #115

Integer:

1. Look over here and apologise.

You've looked and it's good. You, on the other hand, are apparently lazy:

Calculating the correlation without prior logarithm

Your method of finding QC is marked in blue (nothing against it). If you look carefully, you'll see that inverting one of the BPs without prior logarithm changes the absolute value of QC. If you think the distinction is weak, then try taking other variants of price BPs rather than EURUSD and GBPUSD...

Put in a good Advisors that work Do you need the

Dmitry Fedoseev 2010.10.04 23:29 #116

That's right, 1/X is not a linear transformation of X, the graph is reversed and is very similar to the original, but it's still not an exact copy in reversed form. There may also be errors due to the limited number of decimal places in the double variables. [deleted, slip of the tongue].

Errors, bugs, questions What do you think Indicators with alerts/signal

Sceptic Philozoff 2010.10.04 23:30 #117

hrenfx: If you look carefully, you will see that inverting one of the BPs without prior logarithm changes the absolute QC value.

Explain to me, the obscure one, why | QC | must be invariant with respect to the inversion of the series.

hrenfx 2010.10.05 00:10 #118

Mathemat:
Explain to me, the obscure one, why | CC | has to be invariant with respect to the inversion of the series.

It's hard to explain the obvious. Let me put it this way:

Suppose you need to determine a linear relationship between EUR and JPY relative to USD. Obviously, this relationship does not depend on which BPs with EUR and JPY are available to you. For example, whether you take EURUSD with USDJPY or USDEUR with USDJPY or EURUSD with JPYUSD, the relationship between EUR and JPY (relative to USD) will always be unambiguous.

A linear relationship is characterised by a QC with pre-logarithmic BPs. Examples of the reason:

Taking EURUSD and USDJPY. log(EURUSD) = a * log(USDJPY) -> a = log(EURUSD) / log(USDJPY). Let us remember this expression.
Take USDEUR and USDJPY. log(USDEUR) = b * log(USDJPY) -> b = log(USDEUR) / log(USDJPY) = -log(EURUSD) / log(USDJPY) = -a.
Take EURUSD and JPYUSD. log(EURUSD) = c * log(JPYUSD) -> c = log(EURUSD) / log(JPYUSD) = -log(EURUSD) / log(USDJPY) = -a.

The main thing is to understand that the relation between EUR and JPY, relative to USD, does not depend in any way(the absolute value of the relation does not change) on the available BP. It would be strange if Mathemat had EURUSD and USDJPY VRs and Integer had EURUSD and JPYUSD VRs. And both argued that the relationship between EUR and JPY relative to USD is different...

Thoughts on some of Do you need the Elite indicators :)

Prival 2010.10.05 04:07 #119

hrenfx:

...

Try to avoid algorithms where summ(X)*summ(Y) is present. Here's https ://www.mql5.com/ru/forum/107017/page5 why the results are different. Although the formula itself is correct, rounding errors have this effect

СанСаныч Фоменко 2010.10.05 06:34 #120

hrenfx:

You've seen it and it's good. You, on the other hand, are apparently lazy:

Your method of finding QC is marked in blue (nothing against it). If you look carefully, you'll see that inverting one of the BPs without prior logarithm changes the absolute value of KK. If you think the distinction is weak, then try taking other variants of price BPs rather than EURUSD and GBPUSD...

Wonderful formulas, and most importantly, correct. But where is the proof that they can be applied to forex BPs? EURUSD histogram. Your formulas are to the red line and the reality is quite different.

Zero sample correlation does not necessarily mean there is no linear relationship - page 12