Skip to content

Latest commit

 

History

History
2 lines (2 loc) · 684 Bytes

model-specification.md

File metadata and controls

2 lines (2 loc) · 684 Bytes

Model specification

The Portfolio- and Risk-management engine is based upon a Multivariate Generalized Laplace (MGL) distribution. Because its tails are not exponentially bounded, the MGL distribution belongs to the class of so-called "heavy-tailed" distributions which can accommodate extreme market events. To model different return regimes, multivariate asset returns are filtered using a least-squares filter with L1-penalization. Additionally, the engine models asset volatility as multivariate asymmetric Generalized Autoregressive Conditional Heteroscedasticity (GARCH) process, which allows the model to reflect many of the stylized facts of financial markets accurately.