What statistical methods can be applied to non-random values?

I want the scientific approach to the analysis of relationships between currencies, for example USD/RUR and EUR/RUR.
As I understand it, the exchange rate is not a random variable in the mathematical sense, that is, concepts such as, for example, variance, t-distribution and its properties, correlation coefficient, the criterion Studenta, etc. are in this case not applicable.
Question: what statistical methods can be used to explore dependencies nonrandom variables?
March 23rd 20 at 19:36
2 answers
March 23rd 20 at 19:38
I do not think that asking such questions you can easily find the solution ;) especially in pairs both currencies, their randomness is the most difficult to detect, as those involved the best minds around the world.

Try with something easier, such as cryptocurrencies.

p.s. of ready-made tools, there is a good package weka, not only the interface but the java library.

99% of the work is preparing data suitable for analysis by the selected algorithm. Ie not mindlessly try to feed it algorithms just a stream of quotes.
Well, I'm actually studying crypto exchanges, they are USD/RUR is also traded. I seem to be the attractive pair of Fiat/Fiat because the range of possible values of the course there is much more limited extent than in cryptomonad. And when there is a specified "corridor", it is easier to try to learn something from fluctuations in prices.
Looking for people who share my interese in similar searches. - quinn_McLaughlin commented on March 23rd 20 at 19:41
@quinn_McLaughlin, of course of course, remember a year and a half ago I well traded in pairs, usd/cny on the exchange bitshares ... while liquidity fell through the floor (and then the local economy has not caught a Swan, just a month ago got out), now it is not.

I will repeat an important word - liquidity. In cryptoamnesia to trade in usd/eur/cny/... only in small amounts, and in a few thousand bucks (otherwise you are starting your own orders move the market not in their favor)... and Yes, the currency market is not nimble, fluctuations of 1-3% with a period of a month and a half, it means you can in a month, ideally, to do 3% per month income (realities are quite different).

And Yes, cryptoamnesia not a regulated market, stock exchange there love their client funds as their own, literally. The likelihood of losing or locking your account is very high. While you play the other casaca bucks, you few people are interested, but as soon as your amounts become larger in the order, becoming different, but to complain you have nowhere.

spoiler
8_fent3mjgnt-wnbymyfgr_dogm.png
this chart usdeur, on the basis of btcusd and btceur exchange bitstamp (the oldest and most reliable exchange in cryptoamnesia, I think they have even the trading volumes are not drawn, but this requires verification), it's nice to corallium with the stock market.

You see these peaks - it's your loss. They occur in a moment of panic when chances to lose their money in such moments you can easily lose half of the money (again, no liquidity), and no matter what it happens a couple times a year, this is enough to devour your ready income at once. - adriel.Dic commented on March 23rd 20 at 19:44
March 23rd 20 at 19:40
If you want "Scientific" approach to the issue, then start, though, would be to learn the terminology.
Y=sin(X) - then Y not random variable.
Y~sin(X) and then Y is already random.
The exchange rate is non-random value only in the case that it forcibly installs the Central Bank. Here in the USSR the rate of USD/RUR was definitely not accidental.
And today rates (quotations) of currency - the value is completely random (in terms of Mat.statistics, of course).
And it's time.
All the listed stat.features - and neperechislenie - it is currently applicable to your examples. If you want to get scientific and learn science, which is called "mathematical statistics" to begin with.
That's two.
For the study of "non-random dependency values" (or non-random dependencies of all sizes) are used in the representation of such dependencies as functions studied from the fifth class of secondary school. Well, for example Ohm's Law is not accidental (and even then within certain limits) the dependence of the three values. But that's all it to random values, which are quotes - the relationship is not quite the word.
That's three.
The fact that you want to go "scientific" to the task of Forex is commendable from the point of view of promoting scientific curiosity, but absolutely harmful - from the point of view of worldly expediency. Useful - because "on the road", if you have the zeal and not played will have the chance to study all statistical methods, including artificial neural networks, genetic algorithms and fractal analysis, which in this area are already being applied seventy years. Harmful - because almost no one this task is not decided and it looks like in the next - and not only future decision she will not give in. Why - this is a more profound question.
It is a four.
Thanks for the detailed response.
The question for paragraph 1. If the exchange rate of a currency, the value is random, the distribution of values of this course should approach a normal distribution, right? And it absolutely not close, as I understand it. But since I just started to understand the statistics, something can be wrong.
Question to step 4. Here depends on the formulation of the problem. If for example the aim is to track random the reduction rate (i.e. the situation when two exchanges, the exchange rate has not changed, and the third was considerably lower), then common sense dictates that it must be solved somehow. It is here that the question of applicability of various criteria to evaluate the deviation or as random for one of the exchanges, or as to the systematic value pair as a whole. - quinn_McLaughlin commented on March 23rd 20 at 19:43
@quinn_McLaughlin, If the exchange rate of a currency, the value is random, the distribution of values of this course should approach a normal distribution, right? - no, not true. Moreover tell you, for nonstationary series, which are time series of prices - the concept of "distribution of values" in a direct interpretation is not applicable.

- then common sense dictates that it must be solved somehow. - various methods - from multivariate time series with time lags before analysis of emissions. But to the prediction of exchange rate is irrelevant. - Serenity_Hodkiewicz6 commented on March 23rd 20 at 19:46

Find more questions by tags Mathematical statistics