Estimation of Large Financial Covariances: A Cross-Validation Approach
提出一种新的协方差估计量，通过指数加权平均和交叉验证非线性收缩样本特征值，适用于金融时间序列的异方差环境，在大维度下表现优于现有收缩估计量。
In this article, the authors introduce a novel covariance estimator for portfolio selection that adapts to the persistent heteroskedastic environments of financial time series by employing exponentially weighted averages and nonlinearly shrinking the sample eigenvalues through cross-validation. Their estimator is structure agnostic, transparent, and computationally feasible in large dimensions. By correcting the biases in the sample eigenvalues and aligning their estimator with more recent risk measures, the authors demonstrate that their estimator performs well in large dimensions compared to existing covariance shrinkage estimators through an empirical application in active portfolio management.