Dependency statistics in quotes (information theory, correlation and other feature selection methods) - page 61

New comment
 
faa1947: If the model is good, the errors will be normal. But this is unlikely. Even with a forecast one step ahead, the headaches are over the top.
OK, normal and independent errors add up to about sqrt(n). The growth is there, but not even proportional.
 
IgorM:

Hmm, why the rules on Onyx and ask here?

ZS: I'm already used to this forum - here some participants can scatter their bright thoughts on a dozen of topics, collected by search in one place, but to collect your posts on all runet...


It's not mine, I'm just a consumer, of which there are plenty on the net.

There's a presentation on onyx, said for those interested. The author left there about 5 years ago. Neither the construction, nor the rules, nor the interpretation have changed since then, standing as Everest.

And I'm not questioning here, but suggesting a discussion, as the topic of TI application is stated.

I'm not calling to collect my thoughts, there's no need for that.

 
...:
Ok. Then why is econometrics about predicting the next bar/step? Permanently - only one bar/step. What if the forecast is for two or more bars? And when speaking about a forecast for one bar, does it mean a specific mathematical model or, in principle, any forecast states that beyond one bar its feasibility gets closer to zero with each successive step?
See above. I have answered twice. If it is not clear, I am ready to clarify.
 
VNG: And I'm not asking a question here, I'm suggesting a discussion, as the topic of applying TI is stated.

OK, Nikolai, what is it that you don't understand in the top-starter's first post? What repulses you that you do not accept?

Well, I'm tired of answering questions in which there is no question at all...

 
Mathemat:
Well, okay, normal and independent errors add up to about sqrt(n). The growth is there, but not even proportional.

That's the thing, they are not normal, but God forbid stationary, or not quite stationary, dependent or not quite dependent..... On top of all that, there are also errors in the estimated model parameters.

But this is specific, not available when trading on paternals.

 
faa1947:

That's the thing, they are not normal, but God forbid stationary, or not quite stationary, dependent or not quite dependent..... On top of all that, there are also errors in the estimated model parameters.

But this is a specific not available for trading on paternals.

So here comes the bullshit, which no econometrics can cope with for some reason.

The distribution is unknown, the ACF is unknown, nothing is known at all. And how does econometrics calculate anything else?

I have long ago stopped estimating the statistical parameters of bars, it is useless.

 
VNG:

And I'm not asking a question here, I'm suggesting a discussion, as the topic of applying TI is stated.

Let's discuss, there is interest among the audience, but if you want to have a serious discussion, you need to thoroughly collect information for discussion - a new topic and in the first post either discussion material or links
 
HideYourRichess:
I was reading one of the "turtles" the other day, the one who is a former programmer. He says seemingly trivial things, but very interesting. He divides the approach to trading into intuitive and emotional. He says it's different. Emotional - is the same 90-95% of plunderers, while the intuition, which is based on vast experience - this is exactly what he preaches. He writes that there is a lot that simply cannot be formalized, while the eye-brain perceives everything very clearly.
In this regard, the notion of "gut feeling" needs clarification.

It is unlikely that such concepts can be refined in a quantitative or algebraic sense.

Although in chess, machines have "learned" how to beat grandmasters who use both intuition and flair.

 
Mathemat:

So here comes the bullshit that no econometrics can deal with for some reason.

The distribution is unknown, the ACF is unknown, nothing is known at all. And how does econometrics calculate anything else?


It is not known in TA (paternals), because there are no such concepts there. But in etonometrics everything is known at the level of numbers. Problems are identified. We know what can be done, and what we don't know how to do yet. And just because we don't know how to calculate some numbers, it doesn't mean they don't exist. A year ago I showed that indicators should not be used head-on at all. I showed a year ago that head-on indicators should not be used at all. But people use them. They say: "Well, we can see!
 
faa1947:
See above. I have answered twice. If it is not clear, ready to clarify.

It's not clear.

I am interested in two things:

1. does econometrics treat the forecast as a price/time coordinate, or e.g. price/open date - what does level forecast mean? Does econometrics consider a forecast not a point/level but a trend change?

2. A particular forecast model or generally any forecast from an econometric perspective decreases at each successive step?

And more details for the layman, please ;)

1...545556575859606162636465666768...74
New comment