Matstat Econometrics Matan - page 28

New comment
 
Maxim Kuznetsov:

Forex is the bottom of the world's monetary and currency system. Where there is a shortage or a surplus is dumped here

Why do you call it the bottom?) Even a cesspool is called a septic tank)) And this is the bottom)) If it is drained, then drain tanks)) With strict accounting and trading rules)))

 
Comrades, please stick to the topic of the thread)
 
Aleksey Nikolayev:

The point is that the same 'real-world' data can be interpreted under completely different models and assumptions. There is even the notion of a "bouquet of models". Accordingly, correlation, as an abstract concept, can mean different things depending on the specific model used. That is, there is no clear correlation "data => correlation".

For example, correlation is undefined if the model used is that the data are deterministic - all covariances and variances are zero and a division of zero by zero occurs.

Within econometric models, meaningful use of the concept of correlation occurs in the form of the concepts of ACF and partial ACF.

Thanks for the answer, but it is still not clear what the knowledge of "true" correlation, which does not happen in practice, gives us for practice)

 
secret:
Thanks for the answer, but it is still not clear what the knowledge of the "true" correlation, which does not happen in practice, gives us for practice)

In short, it allows you, for example, to calculate the confidence interval for the correlation coefficient, or how significant it is when it differs from zero. Any normal statistical package does this calculation when checking for correlation.

PS. The link to San Sanych's article is just right)
 
Oh, that's more interesting, thank you).
But I still don't see much benefit. It is already clear that the estimates will be very imprecise. And with the correlation it is also clear that values < 0.7-0.8 mean nothing. It's the stability of metrics, not their accuracy, that matters in the market.
 
secret:
Oh, that's more interesting, thanks)
But I still don't see much benefit. It is already clear that the estimates will be very imprecise. And with the correlation it is also clear that values < 0.7-0.8 mean nothing. In the market the main thing is stability of metrics, not their accuracy.

Well, stability can probably be calculated somehow).

If you trade by hand and by eye only, you probably don't need it all. But when all this needs to be somehow put into a robot, sooner or later there is always a question like "how much in grams? )

All modern devices with "brains" are stuffed one way or another with matstatic algorithms, for they must work more or less autonomously in a changeable environment. For example in network card, thanks to which we access the internet, algorithms based on the matstat of Markov chains etc. can be used. With trading robots it is much more complicated, but in general it is more or less the same.

 
My trading robot takes 10 lines)
 
I do.
 
Aleksey Nikolayev:

All modern devices with "brains" are in one way or another stuffed with matstatic algorithms, because they have to work more or less autonomously in a changeable environment. For example, a network card, thanks to which we access the Internet, may well use algorithms based on the matstatic of Markov chains, etc., etc. With trading robots everything is much more complicated, but in general approximately the same.

on the physical level - yes, only mathematical models describing physical processes

the network card can do little to control the robot's processing.

the network card can do little, probably more correctly as an example - the ISP linking equipment or SDH

 
secret:
My trading robot takes 10 lines)

On R?

1...212223242526272829303132333435...38
New comment