How to minimise index correlation - page 6
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 you have to look for a point where the correlations are minimal
> (ideally zero, but it won't be zero) and so with each
> recalculate that point...
You seem to be dreaming of the "SSA" method.
It looks for orthogonal components, some of which are trending, some of which are oscillatory.
Only... a shift of even 1 bar can cause the components to change significantly
for no apparent reason and there's a shift in the reference frame...
That's a lot of redrawing of the whole window.
+ there will be people's favourite "edge effect".
> so you have to look for a point where the correlations are minimal
> (in the idial it is zero, but it will not be zero) and so with each
> new counting recalculate this point
Sounds like you are dreaming of the "SSA" method.
It looks for orthogonal components, some of which are trending, some of which are oscillatory.
Only... At shift even by 1 bar components may significantly change
for no apparent reason - and there will be a change of reference system -
and it will change the whole "window".
+ there will be the people's favourite "edge effect".
You obviously know more about them than I do, as well as about filters, maybe you will come up with something new, but I doubt it.
From my primitive viewpoint, I believe that marginal effects cannot be eliminated by interlaying a row and mixing the same row, and this effect occurs when the row is shifted (or not? I'm confused).
It seems to me that there is a way around this problem. Maybe I'm wrong, I'm trying to check, but with my "perfect knowledge of mathematics and excel")) it's difficult, the extrapolation is not working with the absolute values of the spectra of the series at decomposition.
I will post it on the forum as soon as I think of something sensible to say in words.
but at the slightest movement the edges flap like wings.
I wrote down the reasons for this - the base changes by leaps and bounds.
By the way, I also do not really understand ;-).
> it seems to me that there is a way round this problem
Hmm... somewhat... an oblique basis?
I understand an edge effect as when the row in the middle is more or less stable-
but at the slightest movement the edges flap like wings.
I have written down the reasons for this - the basis changes by leaps and bounds.
By the way, I also do not really understand ;-).
> it seems to me that there is a way round this problem
Hmm... somewhat... an oblique base?
what kind of basis? )))) it's hard for me to say in mathematical language. there are no such methods in common access, so there are no terms for this method, and terms from standard methods are not very suitable, people are also surprised how is it that you can not say in words, if it were so easy to describe through conventional methods, there would be no problems.
In my case, it's a real bummer: I can't easily describe old methods of any more complex mathematical analysis in words, I have to remember them. But here is something new, and also in my head, which is not so easy to understand and is presented by figurative thinking from scratch. ))))
I do not know what is there and how it is counted. I look at the process in development, as a multiplier in my head, and I do not know how to scroll through formulas, you need to see not just a formula, but what values it will produce when you substitute different coefficients to see the physics, the dynamics of the process.
Although I may be mistaken and inventing the same bicycle, but with a different painting)))) do not know, dig further
this is what is a baseline in your understanding? relative to what? and what is an oblique baseline?
is this what you mean by a baseline? in relation to what? and what is an oblique baseline?
It is a system of orthogonal functions. There is nothing to it.
An oblique basis is when the basis was first orthogonal,
and then it's not really orthogonal and it's not really a basis,
and we use it a little bit like this... but, like, we know when to stop.
I really haven't tried it that way.
And does it even make sense to say that the length of a row of basis functions
is the length equal to the width of the window... i.e. before using that basis there
You have to extend it somehow.
But taking the basis from one window and applying it to another window...
that's a slightly more interesting option. I wish somebody would tell me the meaning of this operation.
I would like to remind that the founder of correlation analysis K.Pearson introduced the concept of false correlation in connection with measuring the closeness of relationship between two relative quantities (indices) Z1=x1/x3 & Z2=x2/x3, where x1,x2 and x3 are uncorrelated!
it is also useful (IMHO) to read the attached paper...
;)
Basis?
It is a system of orthogonal functions. There is nothing to it.
An oblique basis is when the basis was first orthogonal,
and then it's not really orthogonal and it's not really a basis,
and we use it a little bit like this... but, like, we know when to stop.
I really haven't tried it that way.
And does it even make sense to say that the length of a row of basis functions
is the length equal to the width of the window... i.e. before using that basis there
should be somehow prolonged.
But taking the basis from one window and applying it to another window
is a slightly more interesting option. I wish somebody would tell me the meaning of the operation.
may I ask, as a more experienced person.
1- Do I understand correctly that by solving the edge effect problem within certain reasonable limits, we are happy?
2- In essence the edge effect manifests itself as an unpredictable jump at the ends of the sample when the sliding window is shifted by a sampling unit.?
3- If we cannot predict this jump completely, is it sufficient to
a) know the direction of the jump
b) the magnitude of the jump (or something like the scale of the jump)
c) we need to know both a) and b)
I would like to remind that the founder of correlation analysis K.Pearson introduced the concept of false correlation in connection with measuring the closeness of relationship between two relative quantities (indices) Z1=x1/x3 & Z2=x2/x3, where x1,x2 and x3 are uncorrelated!
it is also useful (IMHO) to read the attached paper...
;)
But thanks anyway, it's better to consider such points in calculations at once.
By the way, there is a branch about correlation - correlation with time shift. this is about the same thing I'm trying to talk about.
may I ask as a more experienced person.
1- Do I understand correctly that by solving the edge effect problem within certain reasonable limits, we are happy?
2- In essence the edge effect manifests itself as an unpredictable jump at the ends of the sample when the sliding window is shifted by a sampling unit.?
3- If we cannot predict this jump completely, is it sufficient to
a) know the direction of the jump
b) the magnitude of the jump (or something like the scale of the jump)
c) we need to know both a) and b)
2 - yes.
3 - I don't know.
But thanks all the same, it's better to take such moments into account in calculations.
By the way, there's a thread on correlation - time-shift correlation. it's about the same thing I'm trying to talk about.
I'm retarded by nature (sort of with a lag like MA;) - but it's just impossible for me to understand what you're talking about.
Sometimes there is a suspicion of high-level trollishness - muddying the waters without a fucking clue, neither in the subject area (in terms of trading), nor in data analysis methods...
At least the spectrum has been sorted out - our harmonious one?
;)