The St Petersburg phenomenon. The paradoxes of probability theory. - page 13

New comment
 
Aleksey Nikolayev:

Naturally, from a single realisation of a random process (our price series) we cannot establish its exact form. This statistic will only allow us to determine the probability of being wrong, assuming the process is other than a random walk (a Wiener process). If this probability is less than the threshold we set - we assume the process is distinct from a random walk - this is the standard matstat approach. The difference from a random walk is interesting for us because the difference is a necessary (although not sufficient) condition for possibility to trade.

It is interesting conversation, as on a lounge, we sit together, in the evening and everyone tells the story, really, it is pleasant to read such conversations.
There are many statistics where we look for a process deviation from a random one, there are many examples, in particular the Hurst Index.
 
Yuriy Asaulenko:

I went to his seminar about five or six years ago. He was talking about his system and about long tails. A.G. did not have anything and never has. Now he is head of the investment department at Finam, and generally out of the game, like all bosses.) Now you can remember the system and the fights they were fighting.

You may be right, but his theory looks quite sound and worthy of study and testing.

 
Yuriy Asaulenko:

Judging by the description, there is, and lots of it, and not just in Python.

The question here isn't even... is there something else like...?, but what exactly is needed? What is missing? One chooses from that, not from something exotic somewhere.

The goal is simple - get it all together so you don't have to choose. Like that movie about Munchausen - "and then we decided to combine")

 
Yuriy Asaulenko:

Python, in terms of statistics, is much more interesting and better than R, and, imho, much easier and faster to do everything. Take your word for it. If you want consistent Python (so the libs don't fight) - put Anaconda and work on Spyder.

All this is for geeks. I don't need all the extra features.

I use google colab to do some calculations in python right from the browser without installing anything, mostly just to learn about MO

I don't have any real problems...
 
Novaja:
Interesting conversation, like in a lounge, we sit together in the evening and everyone tells his story, really, it is pleasant to read such conversations.
Now about the sore point, we are kicking the same stone, only each from his side, there are a lot of such statistics, where we are looking for deviation of a process from a random one, there are many examples, the Hurst index in particular (sorry for triviality) how will your statistics be better?

I'll start by answering with a question) Do you know the distribution law of the Hurst index for a random walk (Wiener process)?

 
Aleksey Nikolayev:

To start with a question) Do you know the Hearst distribution law for a random walk (Wiener process)?

Tell me))

 
Maxim Dmitrievsky:

I use google colab to calculate something in python directly from my browser, without installing anything, mostly just to learn about MO

Checked it out. I couldn't even upload a file from my computer.)

Actually, Jupiter is in Anaconda, but spyder is more interesting and has more features in terms of development environment.

 
Aleksey Nikolayev:

First I will answer with a question) Do you know the Hearst distribution law for a random walk (Wiener process)?

If you take a barometer that shows the weather, small variations in one direction or the other, say within normal limits, you won't even feel that change in the weather.

 
Aleksey Nikolayev:

The aim is simple - to get everything together so you don't have to choose. As in that film about Munchausen - "and then we decided to combine").

To have it all together - there's no such thing.))

 
Yuriy Asaulenko:

To have it all together - there's no such thing.)

What about strength in a bunch?

1...67891011121314151617181920...25
New comment