A0746
Title: Augment large covariance matrix estimation with auxiliary network information
Authors: Shuyi Ge - University of Nankai (China) [presenting]
Abstract: The aim is to incorporate auxiliary information about the structure of significant correlations into the estimation of static high-dimensional covariance matrices. With the development of machine learning techniques such as textual analysis, granular linkage information among firms that used to be notoriously hard to get is now becoming available to researchers. The proposed method provides an avenue for combining those auxiliary network information with traditional statistical regularization models, mainly banding and thresholding, to improve the estimation of a large covariance matrix. Simulation results show that the proposed adaptive correlation thresholding and banding methods generally perform better in the estimation of covariance matrices than competitors, especially when the true covariance matrix is sparse and the auxiliary network contains genuine information. Empirically, the method is applied to the estimation of the covariance matrix of asset returns to attain the global minimum variance portfolio.
