ForumSections

The Sultonov Regression Model (SRM) - claiming to be a mathematical model of the market. - page 12

New comment
Avals:

It is not a question of how or what to base a prediction on, but how to check its validity. If the residuals (error) are not Gaussian distributed, it's no good).
Let's process real BPs and judge, with a decent description of history first, before making predictions.
Avals:

No, the residuals are tested for normal distribution (z-test for example). Stationarity is you test for something else)))


Even double-checked.

No. Taking EViews.

Normality test as a reference (Jarque-Berg) among other descriptive statistics.

Then there is the unit root test: 6 types of test with 8 types of automatic lag length selection; for level, 1st difference, 2nd difference; with inclusion of bias, trend and offset and without anything.

The normality test is a test of nothing.

It is more reliable to perform the test after detrending and removing the cyclic component.

Avals:

It's not a question of how or what to base a prediction on, but how to check its validity. If the residuals (error) are not distributed according to gauss, it's not good))


Why ?

Does this mean that something else can be taken out of the error part or is it indicative of randomness in the error-free part ?

yosuf:
Then describe how you want to get "deterministic", and without noise?

You take a waving machine. The remainder is noise. What is the noise in the dash?
Demi:
it means that the variables are deterministic rather than random
OK, moving on - logically such a model should predict interventions )
faa1947:

You take a mashka. The rest is noise. What's the noise in the wrecking ball?
In trading, sometimes noise is more important than the wad, I think.
TheXpert:
OK, moving on - logically such a model should predict interventions )
Perhaps also a preliminary market reaction to a news release.
TheXpert:
OK, moving on - logically such a model should predict interventions )

No, of course it doesn't.
Yusuf, excuse me, but you have some kind of ego problem bordering on megalomania. And you've already named the model after yourself and put mystical powers on it. What do you actually have? Regression, that's all.
Mischek2:


Why ?

Does this mean that something else can be taken out of the error part or is it indicative of randomness in the error-free part ?

Yes, it means that something else needs to be taken out of the error part. Otherwise the magnitude of the error is unpredictable and who knows how to assess the accuracy of such a prediction. I.e. the prediction will be like: X +- x.z.)
1...5678910111213141516171819...47
New comment