High-Dimensional Covariance Estimation and Portfolio Selection (ADDED FEE)

• Activity Number: 214073 - High-Dimensional Covariance Estimation and Portfolio Selection (ADDED FEE)
• Type: Professional Development
• Date/Time: Tuesday, August 1, 2017 : 8:30 AM to 5:00 PM
• Sponsor: ASA
• Author(s): Mohsen Pourahmadi*
• Companies: Texas A&M University

Abstract

The course provides a broad introduction to covariance estimation for high-dimensional data and its role in portfolio selection in finance. It is known that in high dimensions the sample covariance matrix is a notoriously bad estimator of its population counterpart. We discuss the details for two useful and viable alternatives to the sample covariance matrix: First, the class of shrinkage estimators, which makes minimal structural assumptions on the population covariance matrix, shrinks the sample eigenvalues toward a central value. It includes the well-conditioned Ledoit-Wolf estimator and is pretty much in the spirit of the ridge regression. The second class is centred around factor models and principal component analysis (PCA) and makes hard to verify structural assumptions. Various case studies and data sets will be discussed in detail using some existing packages in R.