Zero sample correlation does not necessarily mean there is no linear relationship - page 3
You are missing trading opportunities:
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
Registration
Log in
You agree to website policy and terms of use
If you do not have an account, please register
Here's another link at the bottom that says
https://ru.wikipedia.org/wiki/Корреляция#.D0.9A.D0.BE.D1.8D.D1.84.D1.84.D0.B8.D1.86.D0.B8.D0.B5.D0.BD.D1.82_.D0.BA.D0.BE.D1.80.D1.80.D0.B5.D0.BB.D1.8F.D1.86.D0.B8.D0.B8_.D0.9F.D0.B8.D1.80.D1.81.D0.BE.D0.BD.D0.B0
If X,Y are independent random variables, then R(x,y)=0. The converse is generally incorrect.
This is just to confirm the topic of the thread. yes you are right.
A game of numbers.
Addressed to Voznesensky - but applicable to us.
It turns out I was a fool to double-check my code 10 times before I posted it. I looked through textbooks. I checked with matrix samples with known matrix packages. In particular, there's a built-in function in matcadet. I checked it all coincides. But it turns out wrong ...
Maybe you can enlighten me on the right way? Before I'm really wrong.
just in case https://ru.wikipedia.org/wiki/Автокорреляционная_функция
Your link to wikipedia gives the correct definition of autocorrelation, which has nothing to do with what you have in your indicator.
What is the function in Mathcad?
There is also a variant of the "autocorrelation" indicator. Very similar to yours and also wrong. Something you both count differently and don't understand its meaning.
Your link to wikipedia gives the correct definition of autocorrelation, which has nothing to do with what you have in your indicator.
What is the function in Mathcad?
There is also a variant of the "autocorrelation" indicator. Very similar to yours and also wrong. Something you both count differently and don't understand its meaning.
and what are the differences? give me a correct calculation. then we'll talk. so far it's just a blanket statement.
1. pierson is wrong.
2. spearman is wrong
3. ACF is not understood at all
4. you need to understand correctly what correlation =0 means
P.S. write it, it's interesting ... terribly interesting ...
Your link to wikipedia gives the correct definition of autocorrelation, which has nothing to do with what you have in your indicator.
What is the function in Mathcad?
There is also a variant of the "autocorrelation" indicator. Very similar to yours and also wrong. Something you both count differently and don't understand its meaning.
Read that code that Sergei has... https://www.mql5.com/ru/code/8295 Bumped by a few facts:
1. SKO is calculated for the whole sample, i.e. if we set history = 2000 bars, then ACF around the 1500th bar is normalized for the future.
2) Linear regression - also calculated on the entire history
3. the "ACF calculation" comment - a reference value m = [0 ... History] is taken,
and from it the ACF( f( i ), f( i + m ) ) is calculated into the past. I.e. the final graph
is a plot of ACF "by itself with a shift" = [0 ... History].
.
That is, when we look at the chart of this indicator, at the 500th bar it is not the value of window ACF for the 500th bar,
but the normalized-to-future ACF with offset = 500. Only calculated not by sample size = History,
but by sample size = History - 500. And at the 1000th bar, for example, there is a peek in terms of regression and RMS into the future already at 1000 bars,
and the sample "becomes thinner" by 1000 elements. On the zero bar it is clear - a fair one because it is on itself.
.
In general, the indicator has a high-level protection from compilation, although it is in the source code ;-). The protection
is that it shows something, but even if you know what it is, there are a number of incorrect
'architectural' decisions which incline one to just rewrite everything.
.
Hrenfx, thanks for pointing that out. I read the indicator code like a fascinating detective story. Write more :-).
.
P.S.: correct me if I'm wrong.
.
P.S. 2: I don't know how... But it would probably be cool to see a graph of the correct ACF,
plotted against X=bar, Y=value of ACF, and Z- bias between samples ;-)
.
P.S. 3: ACF is normalized to ACF[0], in which I ended up with number = 1440.
Theoretically... if you normalize ACF to sample = 1440, that's fine, but the point is,
that further on in the history the number of bars decreases => normalization presses the graph to zero.
P.S. 2: I don't know how... but it would probably be cool to see a graph of the correct ACF,
plotted against X=bar, Y=value of ACF, and Z- bias between samples ;-)
What's stopping it?
Then again, you have to interpret this graph later... I have to make a hypothesis...
To make a decision-making block, an automatic tester, then a tester with OOS...
By the way, P.S. 4:
- I made a joke in my post about "protection". Mistakes, of course, happen to everyone.
Therefore, it is exactly a joke.
Well... I don't know how :-). I use Mql and C++. And I don't know how to make such nice charts ;-).
Then again, you have to interpret this graph later... I have to make a hypothesis...
To make a decision-making block, an automatic tester, then a tester with OOS...
In csv and excel to chart it. Just to get a glimpse of it.
In csv and in excel to make a diagram. Just to get a glimpse of it
:-)