Zero sample correlation does not necessarily mean there is no linear relationship - page 9
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
A very interesting discussion! Informative and entertaining.
But I have a question - on almost every page of this thread (and in most other threads) the expectation calculated for a forex price series comes up through the post.
And how do you calculate the MO for a non-stationary and non-transcending time series? After all, the concept of MO is only introduced for perfectly convergent BPs.
But I have a question - on almost every page of this thread (and in most other threads) the expectation calculated for a forex price series comes up through the post.
And how do you calculate the MO for a non-stationary and non-transcending time series? After all, the concept of MO is only introduced for perfectly convergent BPs.
This question is actually not that simple. When we say that a series is non-stationary, what some of us actually mean is that this non-stationarity is specific. It doesn't quite fit the traditional framework, it's piecewise, and that's not its only feature. But that's a separate topic altogether, and a very big one.
However, the way the author did it, see pic in the topikstart - certainly not quite right to do so. That's why, yes, there MO, variance and other statistical attributes don't exist, in a statistical sense. But as a result of some mathematical actions, - yes, we can add up monkeys and points piece by piece. :) What the result will be is probably a fable.
so what if there is no linear relationship between the two random processes?
Suppose there is a measure of the linear relationship between the two rows, well, what do you want to
do you do with it? Build a regression? Okay, let's build one and get a good approximation
of one series to the other and then what? Did we get a prediction? Not at all.
And with pirson and correlations it cannot be done!
After all we need dynamics! And what the hell are the dynamics in a scalar quantity,
like correlation? And Pearson's no good at all because it's symmetrical.
we see a symmetric graph and we need an asymmetric measure because one
one series has to outrun the other.
:)
As for non-stationarity and detecting the relationship of one row to another, the
the main thing is to understand what stationarity is and what it's for...
Stationarity is the immutability of moments in time, across the time axis of counts,
even if you start shifting one series relative to another you won't find the connection, because with your
you've destroyed that dynamic relationship by making the series homogeneous in time.
So the main thing is not to overdo it :)
A very interesting discussion! Informative and entertaining.
But I have a question - on almost every page of this thread (and in most other threads) the expectation calculated for a forex price series comes up through the post.
And how do you calculate the MO for a non-stationary and non-transcending time series? After all, the concept of MO is only introduced for perfectly convergent BPs.
In the next thread I proposed to estimate parameters of BP using quantiles. When applied to the expected payoff, it is its estimation by the median (and not the mean) of the last values. The results, by the way, are much better in terms of ROS.
My opinion about stationarity of the series is justified in the same place: at the moment I am working on the model, in which the resulting series is the sum of two stationary (or rather quasi-stationary - with slowly floating characteristics) random processes - Gaussian "calm" and Poisson's "disruptions". With this approach, non-stationarity of the resulting process is apparent. The developments that I have so far allow me to say that I am moving in the right direction on the way to building an approximation to the reality of the price model.
I'd like to insert my five cents.
Correlation is simply a number which is always obtained as a result of a formula calculation, to everything from -1 to +1. Applying the formula to BP we will always get some correlation value: between any currency pairs, between a currency pair and the movement of Jupiter, between a currency pair and anything else.
Any statistics textbook warns of correlations and pseudo-correlations. To separate one from the other you have to use other statistical methods to avoid falling into the delusion of false knowledge. The most effective way to separate correlations from pseudocorrelations, however, is to make some meaningful assumptions about the possibility that two random variables are related. With a priori assumptions about the relationship, we can calculate correlations between the euro dollar and the British pound, but we will never include the Chinese yuan. Using the mataparatus we test these assumptions and even having them it is very easy to obtain pseudo-correlations just because statistical characteristics of BPs, such as the normality requirement, have either been omitted or incorrectly considered.
Privalov wrote a script, the results match the Matkadian ones. I wrote the script without looking at the others, and compared the results - they are the same as Beer's and Matkad's. One hundred and fifty people have already written this QC a hundred and fifty times - and all the results are the same. So why would everyone suddenly rush to rewrite their programs, suddenly finding out that someone has their own interpretation of Pearson's QC?
How do you calculate the IR for a non-stationary and non-collapsing time series? After all, the concept of MO is only introduced for perfectly convergent VR.
And you can't do it with piersons and correlations!
Because we need dynamics! And what the hell are the dynamics in a scalar quantity,
like a correlation?
There is a definition of Pearson's QC on the sample. I can't figure out what everyone is counting by calling it Pearson's QC. Please name the function in Mathcad. So that I can see its description in detail in help.
Laziness is a great power.
http://books.google.ru/books?id=or8sS60Q0ooC&pg=PA159&lpg=PA159&dq=mathcad+corr&source=bl&ots=X2IwE5hGCI&sig=wXlaqOs7LlEQnozhuXyKb2_Fx14&hl=ru&ei=Z4apTPnRRHMyhOvH_1Y8M&sa=X&oi=book_result&ct=result&resnum=2&ved=0CBoQ6AEwAQ#v=onepage&q=mathcad%20corr&f=false
corr(A,B) - calculation of Pearson's sample correlation coefficient.