Zero sample correlation does not necessarily mean there is no linear relationship - page 28

 
hrenfx:

I don't understand what the Bomba is about?

  1. I take any interval.
  2. On it I build a curve, which has the same QC with all the majors.
  3. I take another interval.
  4. On it I build another curve, which has the same CC with all the majors.

So you don't actually have a continuous series of generated data?

Instead you propose to analyse a set of curves?

And call them "synthetic"?

If there is a synthetic that has the same correlation coefficient with all the majors, what does that suggest?

;)

 
FreeLance:

So you don't actually have a continuous series of generated data?

Instead you propose to analyse a set of curves?

And call them "synthetic"?

I'm not suggesting anything. You can find a curve (vector of weights) for any interval that will have the same QC to all majors.

When you change the interval, e.g. by shifting the window, the vector of weights will be adjusted slightly.

This is a matrix problem that can be solved for any BP, including SB. So don't see it as something marketable.

Just pulled the formulas onto the graphs using this twisted method and asked for an opinion.

 
hrenfx:

I am not suggesting anything. For any interval, it is possible to find a curve (vector of weights) that will have the same QC to all majors.

When you change the intrevalue, e.g. by shifting the window, the vector of weighting coefficients will be adjusted slightly.

This is a matrix problem that can be solved for any BP, including SB. So don't see it as something marketable.

Just stretched the formulas onto the graphs with this twisted method and asked for an opinion.

I went on and on :)

I suggest we work with this fan of curves. visualise it first. For example, by limiting the amount of data in the curve to 8 points. then you can cram it into the indicator buffers.

I'd be curious to see.

Would you be so kind as to make one?

 
FreeLance:

Would you be so kind as to create one?

It's better to build in Mathcad. There is no limit to the visualisation there either.

There is only one BUT, the number of such curves can be any. Which one to choose?

 
hrenfx:

It's better to build in Mathcad. There is no limit to the visualisation there either.

There is only one BUT, the number of such curves can be any. Which one to choose?

I think with maximum correlation there can't be many

then see with the minimum...

And according to the topic's name - with zero coefficient to complete the research.

;)

 
hrenfx:

I didn't want to create a new topic. The question is this:

If there is a synthetic that has the same correlation coefficient with all majors, what can it do?

It could be used to analyse (and possibly predict) the movements of the underlying currency against a basket.

In your case, the dollar, if you refer to USDXXX (XXXUSD) pairs as majors.

Unless, of course, it is a synthetic one calculated according to a constant formula that does not change over time.

Here is an example of such a synthetic: SYNT[i] = Root((C0/C1)[i] * (C0/C2)[i] * ... * (C0/Cn)[i], n+1);

// Here C0 is the base currency, C1 ... Cn - other currencies of the basket

;-)

 
MetaDriver:

Here is an example of such a synthetics: SYNT[i] = Root((C0/C1)[i] * (C0/C2)[i] * ... * (C0/Cn)[i], n+1);

What is Root?
 
hrenfx:
What is Root?

Root. Degrees n+1 in this case. I.e. I gave the trivial geometric mean of a set of pairs as a synthetic.

// Popularly, such a thing is called the C0 currency index relative to the C0..Cn basket.

 

I wrote above that there is a vector of weighting coefficients which is adjusted when the sliding window is shifted.

Of course, we can take a year window and create a synthetic that has the same and maximum possible correlation with all the majors in that year.

But let's not forget what QC is, and that it depends on the sample and its size.

 
MetaDriver:

Root. Degrees n+1 in this case. I.e. I gave the trivial geometric mean of a set of pairs as a synthetic.

// This kind of thing is popularly referred to as the C0 currency index relative to the C0...Cn basket.

There is such a thing.