How do you properly compare two non-overlapping rows? - page 4

 
Dmitry Fedoseev:

Tops up...

You need to make 2 PNB equations for each row and compare parameter values and coefficients.

 
Yousufkhodja Sultonov:

You have to make 2 PNB equations for each row and compare the values of the parameters and coefficients.

Exactly, so exactly. Urgent to the PNB - psycho-neurological hospital

 

Maybe the question should have been"How to superimpose two time series on top of each other in the most correct way".

Two series are 'of the same nature' just one lies in one range of the scale from 0 to 100 the other series in another.

With a different sample to calculate the mean or regression and further calculations the result changes relative to the sample (which is in principle natural).

I thought maybe there is a special correct approach for such purposes.


Anyway, thanks for the answers!

 
Dmitry Fedoseev:

PNB - psycho-neurological hospital

:)))
 
Evgeniy Chumakov:

Maybe the question should have been"How to superimpose two time series on top of each other in the most correct way".

Two series are 'of the same nature' just one lies in one range of the scale from 0 to 100 the other series in another.

With a different sample to calculate the mean or regression and further calculations the result changes relative to the sample (which is in principle natural).

I thought maybe there is a special correct approach for such purposes.


So, thank you all for your replies!

google : Market Invariants.

Suppose a share is trading at 10 rubles. Then the increase in its price by 1 rouble will be 10%. Suppose the stock has grown in a few years to 100 rubles. The price increase of 1 rouble will be only 1%. One rouble for a 10-rouble share and a 100-rouble share are very different things, so investors are focused on percentage yields rather than monetary yields, that is, they are focused on relative values rather than absolute ones.

That seems to be what is needed.

 

There has to be some assumption about the approximate type of relationship between the two series. For example, if the series Xn and Yn are related by the relation Yn ~ K*Xn, then to combine them the second series must be divided by the number K, the logarithm of which can be found as the average for the series log(Yn)-log(Xn). If the type of relationship is different, the transformation will be different. If there is no relationship, it is unlikely that there will be any overlap at all.

 

I'm probably off-topic with some rows, but two Mashas of the same period - one open a - the other close and add them levels then they also overlap on levels.

EURUSDH2

 
SanAlex:

I'm probably off-topic with some rows, but two Mashas of the same period - one open a - the other close and add them levels then they also overlap on levels.

As always off-topic. You have one row.

 
Aleksey Nikolayev:

There has to be some assumption about the approximate type of relationship between the two series. For example, if the series Xn and Yn are related by the relation Yn ~ K*Xn, then to combine them the second series must be divided by the number K, the logarithm of which can be found as the average of the series log(Yn)-log(Xn). If the type of relationship is different, then the transformation will be different. If there is no relationship, it is unlikely that there will be any overlap at all.

Why the complexity? Why not just divide each series by its average?
 
secret:
Why the complexity? Why not just divide each row by its average?

The final formula depends on the type of full formalisation of the coupling, taking into account the way the error (white noise) enters there. My approach corresponds toYp= K*Xp*echr(Qp), where Qp is white noise. This seems to be what is called multiplicative noise. You must have a variant of additive noise.