![MQL5 - Language of trade strategies built-in the MetaTrader 5 client terminal](https://c.mql5.com/i/registerlandings/logo-2.png)
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
No, we'll have to linger here. We must clearly understand at what point fitting the model to the process ( Model = f( Process ) ) turns into fitting the process to the model ( Process = f( Model ) ).
And this happens when we finally say: "This model works, it's almost a process! It has to be fitted!". This always happens because ultimately it is the practical use of models that we are aiming for.
No, I will have to linger here though. We should clearly understand the moment when fitting the model to the process ( Model = f( Process)) turns into fitting the process to the model ( Process = f( Model)).
And this happens when we finally say: "This model works, it's almost a process! It has to be fitted!". This always happens because, ultimately, it is the practical use of models that we are aiming for.
The model is considered to work, for example, if the difference |Model-Process| < eps, where eps>0 is a proximity measure.
One can otherwise, in many different ways, define a working model, e.g. Integral[(Model - Process)^2]dt --> min
It is possible to introduce a velocity error, i.e. to introduce the first derivative of the error into the function to be minimised.
One can then introduce the second derivative of the error.
Then, various combinations thereof, and so on and so forth.
All in all, there is a huge room for creativity here.
IMHO
If the model does not work well, we come up with another theoryOr come up with another model within the framework of the theory. now we observe - a limited number of theories and an unlimited number of models.
It is much easier to come up with a new theory. From a practical point of view, it is likely to be of no use either, especially if cause and effect are mixed up, then to illiterate people it will look quite "plausible" on the surface, like Yusufkhoji's, for example. But the author of yet another nonsense automatically becomes the leader of a sect.
But what's the fun in it? Someone out there received a Nobel Prize for econometrics and he doesn't care whether the theory is correct or not. And someone else, having believed in this very theoretical nonsense, now only screws up.
What's the fun in that? Someone out there got a Nobel for econometrics and doesn't care if the theory is correct or not. And someone else, having believed in this very theoretical nonsense, is now only screwing up.
Here's the textbook, I don't see a single item that got a Nobel Prize. Can you give me a hint?
ARCH/GARCH Eng and Granger for example
ARCH/GARCH Eng and Granger for example
It's nice to deal with literate people.
But I personally don't mess with it in Reshetov's terminology - all useful stuff.
Here's the textbook, I don't see a single item for which they gave a Nobel Prize. Can you give me a hint?
I look in a book, I see a fig?
You would do better to look at other sources, including econometrics, rather than dwelling on nonsense textbooks with already outdated, if not false, methodologies.
For example from wikipedia:
Non-parametric econometrics is a section of econometrics that does not require specification of the functional forms of the objects being estimated. Instead, the data itself forms the model.P.S. The most delicious, I have specifically highlighted in bold, so that you do not have in the future to look for a black cat in a dark room, especially where there is none.
It's nice to deal with literate people.
But I personally don't mess with it in Reshetov's terminology - all useful stuff.
Reshetov:
instead of getting hung up on nonsense textbooks with already outdated, if not false, methodologies.
......Instead, the data itself forms the model.What you refer to as new trends, I paid these trends 30 or 40 years ago. In the USSR, the fields of artificial intelligence and pattern recognition were extremely well developed. I am not interested in it.
The lack of interest among the participants of this forum is explained not by the obsolete parametric methods, but by the most common crude ignorance and a desire to show off. And it is done by reinventing another bicycle.
Flip through this thread and you will see that there is no constructive criticism of what I lay out, as well as no suggestions for the development of the semiproduct I laid out in the thread.
Now, a couple of pages ago we stopped before defining the stability of the model. Maybe non-parametric econometrics can solve this question? What should the model be like to be able to trust the prediction? Just don't do the forward test.
Or broader. What specific problems not solved by parametric methods can at least some non-parametric method solve.