Unit II - Modellization of multivariate financial indicators with techniques of RMT and information theory

Italian version

Lecturer: Michele Tumminello

We present methods and tools for the analysis and filtering of correlation matrices of stock-return time series, and factor models that can be used to simulate data series with complex correlation structure. In particular, we will consider filtering methods based on Random Matrix Theory, and comparison methods based on the Kullback-Leibler divergence and the Wishart distribution. We will discuss the properties of the correlation matrices at the different time horizons log-returns are calculated and over time. Factor models will be presented that can describe some typical structures of correlation observed in financial markets. Finally, we will study how the correlation between investors’ trading activities can be analyzed through statistically validated networks.