Conditional Covariance Prediction in Portfolio Analysis Using MCD and PCS Robust Multivariate Scatter Estimators
Year: 2016
Type: article
Abstract: In this paper we compare two matrix estimators of multivariate scatter – the minimal covariance determinant estimator MCD with a new proposal an estimator minimizing an incongruence criterion PCS in a context of their applications in economics. We analyze the estimators using simulation studies and using empirical examples related to issues of portfolio building. In a decision process we often make use of multivariate scatter estimators. Incorrect value of these estimates may result in financial losses. In this paper we compare two robust multivariate scatter estimators – MCD (minimum covariance determinant) and recently proposed PCS (projection congruent subset), which are affine equivariant and have high breakdown points. In the empirical analysis we make use of them in the procedure of weights setting for minimum vari ance and equal risk contribution (ERC) portfolios.
Authors: Przemysław Jaśko, Daniel Kosiorowski
Institution: Krakow University of Economics
Language: English
Subfield: Statistics and Probability
Field: Mathematics
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W3023706610