Dependency statistics in quotes (information theory, correlation and other feature selection methods) - page 37

You are missing trading opportunities:
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
Registration
Log in
You agree to website policy and terms of use
If you do not have an account, please register
There you go - right from the first page of the thread and models are specific.
Yes, that's one noise. At the output so to speak))
It could also be on the model input. But in general, it doesn't really matter. Noise is something that is not provided by the model. Or rather the assumption is made that it can be ignored.
))) What is a trend? What is it expressed in? In what units? How does it compare to quotes? How does having a trend and quotes get "noise"? Go for it.
Linear in time trend? Elementary))
Well, then the answer won't be hard for you. Go ahead.
Yes!
"The initial signal is a quote, but we also roughen that down to timeframes, or try to squeeze something out of ticks (but that's the lot of the brave)."
Right. If the original signal is a quote, then its coarsening is a bar. A timeframe is time. Do you translate the number into time somehow?
Is it then correct from the definition: "Noise is usually the difference between the deterministic component of the signal and the original signal" implies that - a quote that falls outside the predictive model is noise?
Then is it a correct consequence of what you have said: truncation, cutting off, averaging and any similar ignoring of the whole is allowed for this kind of research?
yes
That is, we have a sequence of ticks as a process of price movement. From it we can select individual ("correct" according to the method) quotes and discard others ("wrong") at the input. And then from the next (future) sequence the ones that fit into the predictive model are selected, and the others that don't fit are discarded? Is this correct now?
For example, there is a sine wave Asin(Bx) + another signal. The signal may or may not be random. We are trying to estimate a model from the total signal, but only take the sine model for estimation. What we have neglected will give us an error in estimating the parameters of the sine model.
It's the same in trading. Price is a mixture of a bunch of different processes, some of which we neglect within a particular model. The neglected is noise.
What is the deterministic component? The sliff of a big batch of something piecemeal over the course of, say, a week is a deterministic component? I don't understand how there can be a deterministic component to a quote.
Andrew, I understand roughly what you are talking about. Noise can be generated by imperfection of the communication channel, for example. In that sense, 99% of the quotes are probably pure signal, another 1% are distortions by filters on the side of the quote providers.
But it is possible that some very big players can be singled out and their behaviour can be taken as a determining factor, and then everyone else's fuss can be perceived as noise. For example, someone decided to increase the quote by 50 points, but in the process of increasing the price moved down by 1-5 points, i.e., small players were pulling it down. So, the signal from the big uncle distorted and as a result, the noise in the system was generated. So it is like this.
Yes!
Yes what? Where's the answer?
Although it's about to descend into the usual dumb rubbish around models I can feel. Not interested.