Econometrics: one step ahead forecast - page 29

 
Debugger:


If you know the function no errors

A random variable and an error are inseparable. If there is no error, it is just numbers.
 
new-rena:
Come on, I'd like your two years on the subject ...


I provided my arguments to back up what I said ... I don't see any counter-arguments other than the finger tic "come on like..." etc etc.
But good luck in general.

IT WILL BE INTERESTING TO WATCH THE BUMMER,

UNLESS, OF COURSE, THE MODERATORS ERASE MY POSTS.

 
faa1947:

For comparability, we take the same model

Here we have:

EURUSD hp1(-1 to -4) hp1_d(-1) hp1_d(-2)

Here HP is the HP filter of EURUSD. Let us take a filter from DX and forecast EURUSD

That's true, but where are the sources of the indicators? I don't really believe that some kind of averaging brings good results, at least with a probability of 0.5. We need to look at the codes of the indicators first, their mathematics
 
Debugger:


I've provided my arguments to back up what I'm saying... I don't see any counter-arguments other than the finger tic "come on like..." etc etc.
But good luck in general.

IT WILL BE INTERESTING TO WATCH THE BUMMER,

UNLESS OF COURSE THE MODERATORS ERASE MY POSTS.

That's fine. Well, you have to check the correctness of the deal size readings on the instruments ))))
 
I have no doubt there are some good inventors here on this forum, but this path is a dead end.
 
new-rena:
That's right, but where are the sources of indicators? I do not quite believe that some averaging brings good results, at least with 0.5 probability. I need to see the codes of indicators for starters, their mathematics

There is no source code for the indicators. There is a schematic model

EURUSD hp1(-1 to -4) hp1_d(-1) hp1_d(-2)

If you reverse it

EURUSD = с(1)*hp1(-1) + С(2)*hp1(-1) + С(3)*hp1(-1) + С(4)*hp1(-1) + С(5)*hp1_d(-1) + С(6)*hp1_d(-2)

Let's evaluate C(i) coefficients and get some synthetic indicator.

Above, the table shows the estimated accuracy of this synthetic indicator for the quote = 97%.

See what happens in our case.

 

Volumes for AUDUSD,EURCHF, EURGBP, EURJPY, USDCAD, USDCHF, USDJPY, EURUSD


What can you see or guess here? (T minutiae)

Well, the weights of currency pairs are clear - the volumes are different (by height) and the percentage ratio is clear

 
new-rena:
What can you see or guess here? (minutes)
If this is from the terminal, it is not volumes, but activity - the number of trades, which may be 0.1 lots, and may be 100 lots.
 

For example since 1999. AUDUSD,EURCHF, EURGBP, EURJPY, USDCAD, USDCHF, USDJPY, EURUSD

You can start to believe since October 2010. I hope it is clear why

 
new-rena:

Forecasting results using the dollar index.

Dollar index quotes downloaded from the terminal - symbol DX

.........

Graph:

Note that the DX quotes go in the opposite direction to the EURUSD quotes

I record the following model by analogy

EURUSD DX_HP(-1 TO -4) DX_HP_D(-1 TO -2)

where DX_HP is the value of the НР indicator

after estimating the coefficients of the equation we get

EURUSD = 1.3375931364*DX_HP(-1) - 3.27636518775*DX_HP(-2) + 2.58060240407*DX_HP(-3) - 0.623668221765*DX_HP(-4) + 0.061086265724*DX_HP_D(-1) + 0.100405219475*DX_HP_D(-2)

Final scorecard:

R-squared is negative. It does not. It is the result of opposing direction of the DX quote to EURUSD

Let's form a quote using the formula: DXM = 1/DX, i.e. let us take opposite values and apply our model to them. We obtain the final valuation table:

Quite an acceptable result. I make EURUSD forecasts on the basis of quotes of the inverse value of dollar index = 1/DX. I get an extension of the result table

Date Value Change Forecast Forecast Error Forecast Error Change Change
Open Open prices at based on in pips based on in pips forecast forecast
eurusd DX on eurusd on DX
2011.11.09 00:00 1,383 NA 2011.11.09 1,3798 56 1,3663 67

2011.11.10 00:00 1,3524 -0,0306 2011.11.10 1,3613 60 1,3742 70 -0,0032 -0,0167
2011.11.11 00:00 1,361 0,0086 2011.11.11 1,3541 59 1,3766 71 0,0089 0,0218
2011.11.14 00:00 1,3778 0,0168 2011.11.14 1,3676 59 1,3673 69 -0,0069 0,0156
2011.11.15 00:00 1,3624 -0,0154 2011.11.15 1,365 59 1,3634 69 -0,0102 -0,0105
2011.11.16 00:00 1,3525 -0,0099 2011.11.16 1,3529 57 1,3627 69 0,0026 0,001






Conclusions:

1. Successful forecasts increased by 1.

2. the magnitude of the predicted bias became more realistic

3. there was an improvement in prediction with an increase in prediction error.