Remembering veterans: Box and Jenkins - page 6

 
faa1947: Either way. Two alternatives.

Agreed. Either it's clear, or one of two ))))
 
faa1947:

Don't. I won't quote the fable.

Only discussing Box and Jenkins with some extensions. With them ACF plays a very important role. I pointed out which one.

As far as the value of the branch is concerned.

I have posted many times about the caution of interpreting test results and forward tests. Only today in response to Reshetov I have realized one very important idea. Testing and forward testing may be trusted only if the residue = kotir - TC is stationary! If this residue is non-stationary, then testing cannot be trusted at all, and no forward tests will help. It seems to me that a branch could have been started for the sake of such a conclusion. Here's Box and Jenkins for you.



You have a strange understanding of the TS and its results. It is not a prediction of a quote for the next move, but a buy/sell of an asset. The results of the system from entry to exit are the increase in quotes for the time in a trade for longs, and the mirror image for shorts. I.e. TS results are the chart of kotirade at certain periods (when the position is open). Therefore, if you subtract system returns from the kotir series, you will get 0 on these segments. MM of course still affects, but these are details

I.e. it has already been written many times that there is no task to predict where the rate will be at discrete points in time t1,t2 etc., and you keep going back to it :)

 
Avals:

Look at my post about the TS of two wagons. Any TS is a set of formulas in which the argument is a kotir. That's the difference between the set of formulas and the quotient is the residual. Either we study this remainder and it is included in the set of formulas, or we turn a blind eye to the remainder, believing that there is no remainder.

You have a strange understanding of the TS and its results. It is not a prediction of a quote for the next move, but a buy/sell of an asset.

It is a prediction first and then a buy/sell. Even if your TS enters/exits on a random number sensor, it is such a decision to enter at random.

I.e. results of TS are chart of quotient at certain parts (when position is open)

The result - profit/loss

I.e. it has already been written many times that there is no task to predict where the rate will be at discrete points in time t1,t2, etc., but you keep going back to it :)

You do not, and the rest of the world do. Don't generalise like that. Box and Jenkins, for example, have been in translation since 1976, and so the anniversary is 40 years old.

 

faa1947:

Trust the test and the forward test only if the residual = quotient - TC is stationary! i.e. it passes the unit root test. If this residue is non-stationary, then testing cannot be trusted at all and no forward tests will help. It seems to me that a branch could have been started for the sake of such a conclusion. Here's Box and Jenkins for you.


It's not very clear what you mean by kotir - TC?

Give a strict mathematical definition, i.e. how to correctly calculate the difference between kotir and TC, and then we can go on. But so far we are talking about nothing, i.e. something that only you understand. I assume it is about the equity curve in fixed lot trading, but it is only my hypothesis.

 
faa1947:

I.e. it has been written many times that there is no task to predict where the exchange rate will be at discrete points in time t1,t2, etc., but you are still there :)

You do not, and the rest of the world do. Do not be so generalized.

There is no need to predict the absolute value of quotient at each discrete time interval.

Give me a link - who trades like that "in the rest of the world"? :)

 
Reshetov:

It is not very clear what you mean by kotir - TC?

For give a strict mathematical definition, i.e. how to correctly calculate the difference between kotir and TC, and then you can go on. So far we are talking about nothing, i.e. something that only you understand.


I gave the example of the mashups above. In EViews it is the button that calculates such a residual.

I will give you the picture again.


 
Reshetov:

It is not very clear what you mean by kotir - TC?

Give a strict mathematical definition, i.e. how to correctly calculate the difference between kotir and TC, and then you can go on. As for now, we are not talking about anything, i.e. something that only you understand.


Yes, he means that TS is a thing that must give a prediction - the absolute value of the rate at any given time :)
 
Avals:

There is no need to predict the absolute value of the quotient for each discrete time period.

give me a link - who trades like that "in the rest of the world"? :)



You don't and it's a matter of taste, idea of TC construction and whatever else. What is fundamental is the non-stationarity of the quotient and what to do about it. Box and Jenkins have made their suggestion. They say that if the initial quotient is non-stationary, then it should be differentiated in that case. Absolutely. In the hope that we get a stationary residual. This is not a fresh idea. But in the future it was developed, it was fleshed out, but the idea of fighting non-stationarity is still alive and in my opinion it has not been solved in full.
 
Avals:

Yes, he implies that the TS is a thing that must give a prediction - the absolute value of the exchange rate at any given time :)

No. Many ideas are known: kotir value, kotir increment, trend, volu, levels .... It doesn't matter.
 
faa1947:

You don't and it's a matter of taste, the idea of constructing a TS and something else. What is fundamental is the non-stationarity of the quotient and what to do about it. Box and Jenkins have made their suggestion. They say that if the initial quotient is non-stationary, then it should be differentiated in that case. Mandatory. In the hope that we get a stationary residual. This is not a fresh idea. But later it evolved, it grew, but the idea of fighting non-stationarity is still alive and in my opinion has not been solved in full.


1. You are setting an impossible task - building a model that will give a prediction of the absolute value of the price at any discrete point in time is impossible. More accurately, a prediction that makes sense for earnings. Incorrect formulation of the problem takes you in the wrong direction.

2. The non-stationarity of the series is not a problem, because there is no need to have a forecast always and in the form in which you make it (point 1). I.e. again the trabble is in the formulation of the problem.