Zero sample correlation does not necessarily mean there is no linear relationship - page 17
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
So why not take the median of, say, the same values instead of the arithmetic mean? For a distribution with thick tails, this estimate is definitely more efficient. The Pearson QC formula would be more complicated (plus it would become non-linear), but it would still be Pearson's QC - or rather, one of its possible estimates!
The median QC is no more complicated than the classical QC. Only instead of addition there is ordinary sorting of BP terms. And the iteration formula is simple.
Another property of the median QC is that unlike the classical QC, the QC does not change with each sliding window shift.
And one more note, when calculating the MCC, the variance is divided by the variance. But the variance, as well as the MO, is not the classical variance but the median one.
I haven't done a comparison of QC and MQC efficiencies on price BPs.
If the point is to find a linear relationship between price VRs, then the original VRs should be prolagarithmic before calculating the QC.
The other case is funds or raw materials. There, the removal of seasonalities and trades is appropriate.
HideYourRichess:
На самом деле, и на форе есть нечто, что можно считать "сезонностью" и "трендами". Просто не все это видят. Особенно через маленький обзорный экранчик ДЦ, с пипсовкой.
This "fashionable" high-frequency trading is what the "kitchen" DCs pejoratively call "pipsing"...
And they fight it in every possible way.
So - not from commissions on deals, but from sinking depo of a "passenger" live.
;)
I think I'm not the only one who has already pointed this out, but I'd still say it's nonsense.
What exactly is nonsense? Before analysing the prices, they should be logarithmic, because analysing the prices themselves is really nonsense. If you are interested in the correlation coefficient, then after logarithm the resulting date should also be standardised.
Another issue is that the presence of correlation does not mean there is a relationship.
If the correlation coefficient is of interest, after logarithmizing the resulting date, it is a good idea to standardise it as well.
Bringing it to the zero MO. Is this what you mean by standardisation?
True, bringing to zero MO (sometimes also variance to one) is not needed to calculate correlation. Because correlation does not change when a constant is added to or multiplied by a constant.
This "fashionable" high-frequency trading is what the "kitchen" DCs pejoratively call "pipsing"...
And they fight it in every possible way.
So - not from commissions on deals, but from sinking depo of a "passenger" live.
;)
What exactly is nonsense? Before analysing the prices, they should be logarithmic, because analysing the prices themselves is really nonsense. If you are interested in the correlation coefficient, then after logarithm the resulting date should also be standardised.
. There is no need to logarithm anything. You have to analyse the prices themselves. As they are. I can understand logarithm when analysing data over many years. This can be understood, with a great stretch, but in the analysis within a year - it is a completely unnecessary operation. It doesn't do anything. Apart from being near-scientific.
. The thing about standardisation is that it's weird. What is it for? For what tasks?
Another issue is that the presence of correlation does not mean there is a relationship.
. This is not another question - it is the main question. After the question about the possible nature of the connection.