Where your hard-earned money comes to die. Home.
|
|
can crypto pairs be modeled as an Ornstein-Uhlenbeck process?
| Odious bistre site | 11/30/17 | | Odious bistre site | 11/30/17 | | Abusive field pisswyrm | 11/30/17 | | Odious bistre site | 11/30/17 | | Abusive field pisswyrm | 11/30/17 | | Odious bistre site | 12/01/17 | | Abusive field pisswyrm | 12/02/17 | | Odious bistre site | 12/05/17 | | Odious bistre site | 12/11/17 |
Poast new message in this thread
Date: November 30th, 2017 1:47 AM Author: Odious bistre site
estimating the parameters of an observed multi-dimensional OU process is typically accomplished via maximum likelihood estimation, but the problem with MLE is that it becomes computationally intractable as dimensionality increases. the goal of statistical arbitrage is to hedge risk (e.g., by taking opposite positions on highly cointegrated pairs of market-traded goods), but the amount of data required for statistical significance grows exponentially with the number of market-traded goods (e.g., with each crypto market).
here's a paper that uses adaptive lasso to model a multi-dimensional OU process subject to a row-sparsity assumption: https://arxiv.org/pdf/1707.03010.pdf
(http://www.autoadmit.com/thread.php?thread_id=3813284&forum_id=7#34803934) |
|
|