Dependency statistics in quotes (information theory, correlation and other feature selection methods) - page 8
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
in this case independence is not required as I understand it, but is precisely the subject of evaluation.
Required. And entropy is estimated as a probability estimate.
Where is it written that independence is required? Appearance of letters in Russian-language text regardless of context (previous letters)?
Many examples of applications of TC, in Russian language, refer to the analysis of the alphabet of Russian, and other languages, as well as to the analysis of words and phrases (word sequences). And all these characters are not a priori statistically independent, and by these examples are estimated mutual information, a value that shows the amount of dependence. So the a priori independence of the quantities being studied is not a prerequisite for the correct application of TI.
Very weak argumentation, on the level of that somewhere read, that somewhere is used, for something there, according to this... Exactly about it the academic dictionary wrote to us, - "For modeling of communication systems such approach is legitimate, as they are intended for error-free transfer on a communication channel of the information represented by any set of symbols. Where, however, consideration of the value and meaning of the information is essential, the quantitative approach is inapplicable. This circumstance imposes essential restrictions on the fields of possible applications of TC. Failure to take it into account led in early stages of development to overestimation of applied significance.
Where is it written that independence is required? The appearance of letters in the Russian-language text regardless of the context (previous letters)?
It's a flaw in Statistics. I use it myself, by the way.
Wonderful. The Statistica package is the only source to go to for data mining. Therefore, TI should be forbidden to use it. And ban your own brain too, because with Statistica you don't need it anymore.
Roman: Alexey, can you tell, whether it is realistic to translate all this pleasure in the code in the direction we are interested in...
A. Sergeev did something similar while translating Sultonov's indicator into code or am I mistaken?
It is quite feasible. I do not see any limits there, but it is possible to do sums and logarithms in MQL4. I don't know what sergeev did. But as far as I know from other sources, the most difficult part of calculations was calculating the gamma function. TI was out of the question.
HideYourRichess: Youhave elementary events, return, are identical to elementary events from TI? [...] Hence the question, what kind of "symbols" do we have in the market?
There is already alexeymosc's answer on this: they are [relative] increments, which can be discretized specifically for this purpose. My final alphabet contains between 15 and 50 characters.
I anticipate the next question, "Is it possible to do such a discretisation without spilling out of the bathtub boy?". And why not? I really don't have a procedure to check if I've done it correctly, but a few checks of extreme and particular cases show that I've made no fatal error. The source with the receiver is there too.
That's what the communication channel is - not so easy to answer. That seems to be the question you are going to kill me with...
There is a supposed answer that may seem heretical to you: it is present time, i.e. the period of time when information from the past is transmitted to bar zero.
HideYourRichess: If you don't involve "economic and other meanings", then what processes are we talking about? A process is a "physical" phenomenon, it has causes and it has consequences. For example, the process of an apple falling on Newton's head. In the application to markets, the process of buying and selling. Where is it all in the returns?
I think you are excessively mechanistic, my dear. A process can legitimately be an information phenomenon, consisting in the generation of returns based on real processes of buying and selling.
HideYourRichess: ter.ver, on which ter.inf. is based, requires the independence of the events in question, or symbols.
Show me the source in which this is asserted. I doubt you will find one.
Where have you seen that terver requires independence - if this very independence is a definable concept in terver? And what do you think Markov chains are? And Bayes' theorems? And the concept of conditional probability in general?
Avals: Alexey, where are the calculations resulting in "distant and practically reliable dependencies"? And what does net returns without volatility mean (how are they obtained, because just returns contain volatility)?
Well I told you I was ignorant of econometrics and hopelessly confused about the concept of volatility...
I don't want to post the calculation code here, it's still a know-how. But I can tell you in private how I made them.
Isn't it clear from the conditions of the problem? The formulation, by the way, is quite classical for ter.ver.
I don't)))) if independence is required, then why such a thing as conditional entropy?
If the sequence of symbols of the alphabet is not independent (for example, in French the letter "q" is almost always followed by "u", and the word "vanguard" in Soviet newspapers was usually followed by "production" or "labor"), the amount of information that the sequence of such symbols carries (and, hence, entropy) is obviously less. Conditional entropy is used to account for such facts. https://ru.wikipedia.org/wiki/Информационная_энтропия
Finally a reasonable diagnosis can be made.
Open it up.
By the way, Statistics does not have such a useful feature as a genetic algorithm for selecting input variables, which NeuroShell has. That is, one product cannot accommodate everything. Matlab, too, has no built-in functionality for calculating mutual information, but a corresponding module has been written, and, incidentally, is in demand.
Mathemat:
I think you are being overly mechanistic, my dear. A process can legitimately be an information phenomenon, consisting in the generation of returns based on actual buying and selling processes.
Mathemat:
Show me the source in which this is asserted. I doubt you will find it.
Where have you seen that a terver requires independence - if this very independence is a definable concept in the terver? And what do you think Markov chains are? And Bayes theorems? And the concept of conditional probability in general?
Sorry, HideYourRichess, but you seem to have gone off the deep end. I no longer know what to discuss with you since you are so insistent on outright nonsense. Your logic of reasoning
Ни Марков, ни Байес не имеют отношения к ТИ. А тер.вер. имеет.
is completely incomprehensible to me.
I will not believe it. Show me the source that states that
The demand for independence is the cornerstone of ter.faith.