From practice to theory and back to practice - page 43

[חולםטרנסצנדר ᨖ]

линейная комбинация двух нестационарных рядов дает может давать стационарный ряд

as long as the series themselves are fundamentally connected otherwise everything is futile

In forex pairs are not cointegrated, but out of more than two it is possible to find a stationary combination

not really

only in a relatively short interval pairs can pretend to be cointegrated and even pass the ADF test and other similar tests

and then they spread out inevitably

His balance curve is stationary and ergodic

it probably wasn't a true graph

For simpler synthetics, fewer assets are used, but they have, for example, small profits from each series of trades, trades are quite rare, and some of them can last for months - spreads and swaps "eat up" all the profits.

this kind of thing is better done with stock indices, but even there may be hanging gaps (spread corridor shifts) for a month or more

[Дмитрий]

transcendreamer:

only over a relatively short interval can pairs pretend to be cointegrated and even pass the ADF test and other similar tests

OK, let's play...

Tell me about "cointegration" and "ADF test and other similar tests" fora "portfolio of more than two assets" (c).

[חולםטרנסצנדר ᨖ]

Дмитрий:

OK, let's play...

Tell me about "cointegration" and "ADF test and other similar tests" fora "portfolio of more than two assets" (c).

on a relatively short interval

[Дмитрий]

transcendreamer:

on a relatively short interval

Tell me about "co-integration on a relatively short interval" and the "ADF test and other similar tests" for "a portfolio of more than two assets" (c).

Is this better?

[Aleksey Nikolayev]

It seems that cointegration is defined for integrated series. Is it already established by someone that currency price series are such?

[חולםטרנסצנדר ᨖ]

Дмитрий:

Tell me about "cointegration ON THE SHORTEST INTERVIEW" and "ADF test and other similar tests" for "a portfolio of more than two assets" (c).

Is that better?

nothing could be simpler.

optimize a portfolio, e.g. using MGC like this

unload these figures into csv and apply them to any statistical packet (e.g. gretl) equipped with ADF or KPSS or another stationarity test

you may choose such an interval, over which the portfolio will be almost indistinguishable from the stationary series

it is also possible to check manually in excel by regression of residues

(I'm too lazy to check it)

I remember there are 3 or even 4 versions of the test in ADF, and you have to deal with lag parameters there as well

[Дмитрий]

Aleksey Nikolayev:
It seems that cointegration is defined for integrated series. Is it already established by someone that currency price series are such?

A long time ago. Lots of material online.

For example

"Conclusion: the hypothesis of a unit root has been disproved; the series of relative closing price increments of the EUR/USD currency pair is stationary.

Thus, we find that the series of closing prices of EUR/USD currency pair isintegrated of the 1st order". (с)

[Дмитрий]

transcendreamer:

Don't be ridiculous

[חולםטרנסצנדר ᨖ]

Дмитрий:

Don't be ridiculous.

very well reasoned )))))

and most importantly on the merits ))))

Conclusion: the hypothesis of a unit root has been disproved; the series of relative closing price increments of the EUR/USD currency pair is stationary.

Well, everyone knows damn well that almost any instrument is delta-stationary (returns or logreturns), but this is not the point

it would be strange if forex was suddenly not DS hehe

[Decromor]

Uladzimir Izerski:

You don't need to show this kind of thing in full view.