Econometrics: one step ahead forecast - page 63

 
Avals:

Yes, in fact, the forecast is based on the price at time t1. The deviation of the price from it is considered.
And how is the error at time t1 calculated? What is the error? At t1? Or is it some sort of period average?
 
And the deviation in modulo...?
 
faa1947:
And how is the error computed for the waving at time t1? What is the error? At t1? or some average for the period?


let us assume that the forecast is made at time t1. The actual forecast at time t1+1 will be the value of the waving at time t1, and at time t2 also the value of the waving at time t1. So the error will be the value of the price deviation: at the moment t1+1 the error will be Close[t1+1]-mask, and so on. The root-mean-square error is calculated.

We consider changes of the error value depending on the forecast horizon for the only purpose - to compare it with a neutral variant (sb) in order to find trendiness/reversibility

 
jartmailru:
And the deviation in modulo...?

Yes, RMS. There will be no sign.
 
faa1947:
Backtracking. There's a price, but there isn't and we're just about to figure it out

What do you mean by "lagging"? There is a certain period, there is an average value of that period, or a scale. How can the average value of a period lag behind? Of course, if you calculate the scale as a moving average and then use it to predict the future, there will be a "lag" from the present, but the average itself does not lag.
 
Avals:


let's say a forecast is made at time t1. In fact, the forecast at time t1+1 will be the value of the scale at time t1, and at time t2 will be the value of the scale at time t1 as well. So the error will be the value of the price deviation: at the moment t1+1 the error will be Close[t1+1]-mask, and so on. The root-mean-square error is itself calculated.

The only purpose is to compare with a neutral variant (sb) for trend/return detection, and the only purpose is to look at the variation of the error as a function of the forecast horizon.

Where do we get Close(t+1), etc.?

Or is it on historical data and we check if there was a trend there?

 
C-4:

What does it mean to lag behind? There is a certain period, there is an average value of that period, or mashka. How can a period average lag behind that period? Of course, if you calculate the scale as a moving average and then use it to predict the future, there will be a "lag" from the present, but the average value itself does not lag.
We take the text of the indicator. A bar comes in. Add up the last T price values and divide by T. The result is written to the last bar.
 
faa1947:

Where do we get Close(t+1)? etc.?

Or is it on historical data and we check if there was a trend there?


Yes, on history.
 
Avals:

Yes, on history.
It all makes sense.
 
Avals:

This means a return - prices tend to go back to the mach.

So what gives?
.
If you could play the convergence--
like, for example, the price broke through the bag up--
we buy the bag - we sell the price - in the end - we're always on the plus side.
.
But there is no such possibility.