On Thursday, Nov. 12th., Matti Heikkonen will give a talk starting at 10:15 at room M1. (Quantum) as a part of the Joint Applied Mathematics and Statistics Seminar
Portfolio risk estimation based on independent component analysis
Estimating the market risk of a portfolio involves modeling extremal changes in the values of the underlying financial positions. Aforementioned changes are typically dependent and follow heavy-tailed distributions, which makes the multivariate normal framework unsuitable. We propose solving the modeling problem by combining independent component analysis (ICA) with time series models. The proposed model is tested using a data set consisting of American stock returns. We also address the calculation risk contributions, and some issues related to high dimensional data sets
All interested are warmly welcome!