A0455
Title: A generalized precision matrix for t-Student distributions
Authors: Karoline Bax - University of Trento (Italy) [presenting]
Emanuele Taufer - University of Trento (Italy)
Sandra Paterlini - University of Trento (Italy)
Abstract: For Gaussian graphical models, the precision matrix, defined as the inverse covariance matrix, is often used to express the dependence relationship between random variables. However, the Gaussian assumption is hardly satisfied in the financial context and therefore using the precision matrix might not necessarily result in a reliable and accurate picture of reality. We introduce a generalized precision matrix to overcome this issue. As fat tails are a well-known stylized fact in many financial time series, we focus on the $t$-Student distributions, pointing out that the behavior between random assets depends not just on the precision matrix but also on additional elements.
