Nonlinear shrinkage estimation of large-dimensional covariance matrices

Abstract

Many statistical applications require an estimate of a covariance matrix and/or its inverse. When the matrix dimension is large compared to the sample size, which happens frequently, the sample covariance matrix is known to perform poorly and may suffer from ill-conditioning. There already exists an extensive literature concerning improved estimators in such situations. In the absence of further knowledge about the structure of the true covariance matrix, the most successful approach so far, arguably, has been shrinkage estimation. Shrinking the sample covariance matrix to a multiple of the identity, by taking a weighted average of the two, turns out to be equivalent to linearly shrinking the sample eigenvalues to their grand mean, while retaining the sample eigenvectors. Our paper extends this approach by considering nonlinear transformations of the sample eigenvalues. We show how to construct an estimator that is asymptotically equivalent to an oracle estimator suggested in previous work. As demonstrated in extensive Monte Carlo simulations, the resulting bona fide estimator can result in sizeable improvements over the sample covariance matrix and also over linear shrinkage.

Published in at http://dx.doi.org/10.1214/12-AOS989 the Annals of Statistics (http://www.imstat.org/aos/) by the Institute of Mathematical Statistics (http://www.imstat.org)

References in corpus (4)

• Regularized estimation of large covariance matrices
• Estimation of high-dimensional low-rank matrices
• Flexible covariance estimation in graphical Gaussian models
• Wishart distributions for decomposable covariance graph models

Cited by in corpus (77)

• Geometric median and robust estimation in Banach spaces
• Estimation of the Global Minimum Variance Portfolio in High Dimensions
• Optimal shrinkage-based portfolio selection in high dimensions
• Optimal Linear Shrinkage Estimator for Large Dimensional Precision Matrix
• Large-Scale Portfolio Allocation Under Transaction Costs and Model Uncertainty
• Testing for Independence of Large Dimensional Vectors
• Improved information criteria for Bayesian model averaging in lattice field theory
• Stable Estimation of a Covariance Matrix Guided by Nuclear Norm Penalties
• Non-linear shrinkage estimation of large-scale structure covariance
• Linear Shrinkage Estimation of Covariance Matrices Using Low-Complexity Cross-Validation
• Estimation of the covariance structure of heavy-tailed distributions
• A tensor-based unified approach for clustering coefficients in financial multiplex networks
• Estimation of the Covariance Matrix of Large Dimensional Data
• Space-Time Adaptive Detection at Low Sample Support
• Analytic solution to variance optimization with no short-selling
• Accurate halo-galaxy mocks from automatic bias estimation and particle mesh gravity solvers
• Covariance matrix filtering with bootstrapped hierarchies
• Bootstrap Based Inference for Sparse High-Dimensional Time Series Models
• Singular Vectors of Sums of Rectangular Random Matrices and Optimal Estimators of High-Rank Signals: The Extensive Spike Model
• Statistical Inferences Using Large Estimated Covariances for Panel Data and Factor Models
• Cleaning large-dimensional covariance matrices for correlated samples
• TPLVM: Portfolio Construction by Student's $t$-process Latent Variable Model
• De-biased graphical Lasso for high-frequency data
• Power Iteration for Tensor PCA
• Random Matrix Derived Shrinkage of Spectral Precision Matrices
• On generating random Gaussian graphical models
• Dynamic Shrinkage Estimation of the High-Dimensional Minimum-Variance Portfolio
• Cross-Validated Loss-Based Covariance Matrix Estimator Selection in High Dimensions
• Optimal Covariance Estimation for Condition Number Loss in the Spiked Model
• Decomposition Pipeline for Large-Scale Portfolio Optimization with Applications to Near-Term Quantum Computing
• Ridge-type Linear Shrinkage Estimation of the Matrix Mean of High-dimensional Normal Distribution
• R-NL: Covariance Matrix Estimation for Elliptical Distributions based on Nonlinear Shrinkage
• On nearly assumption-free tests of nominal confidence interval coverage for causal parameters estimated by machine learning
• Modeling Brain Connectivity with Graphical Models on Frequency Domain
• Estimating Learnability in the Sublinear Data Regime
• A partial orthogonalization method for simulating covariance and concentration graph matrices
• Ledoit-Wolf linear shrinkage with unknown mean
• Irrational Exuberance: Correcting Bias in Probability Estimates
• Estimation of the Kronecker Covariance Model by Quadratic Form
• End-to-End Large Portfolio Optimization for Variance Minimization with Neural Networks through Covariance Cleaning
• Regularizing stock return covariance matrices via multiple testing of correlations
• On the variability of the sample covariance matrix under complex elliptical distributions
• Empirical Bayes PCA in high dimensions
• A shrinkage estimation for large dimensional precision matrices using random matrix theory
• High-dimensional general linear hypothesis tests via non-linear spectral shrinkage
• A comparison of shrinkage estimators of the cosmological precision matrix
• Eigenvalue distributions of variance components estimators in high-dimensional random effects models
• Random selection of factors preserves the correlation structure in a linear factor model to a high degree
• Principal Regression for High Dimensional Covariance Matrices
• Matrix Denoising with Partial Noise Statistics: Optimal Singular Value Shrinkage of Spiked F-Matrices
• Recovery of spectrum from estimated covariance matrices and statistical kernels for machine learning and big data
• Long Random Matrices and Tensor Unfolding
• Adaptive Shrinkage of singular values
• Noise Estimation in the Spiked Covariance Model
• An efficient numerical method for condition number constrained covariance matrix approximation
• A Compound Decision Approach to Covariance Matrix Estimation
• Asymptotic control of FWER under Gaussian assumption: application to correlation tests
• Tests for the weights of the global minimum variance portfolio in a high-dimensional setting
• Application of Laplace filters to the analysis of lattice time correlators
• Sparse sketches with small inversion bias
• Cross-validated covariance estimators for high-dimensional minimum-variance portfolios
• Best approximations of non-linear mappings: Method of optimal injections
• Learning from Forecast Errors: A New Approach to Forecast Combinations
• Fast randomized numerical rank estimation for numerically low-rank matrices
• Analytic approach to variance optimization under an ${\ell }_1$ constraint
• Community-level Contagion among Diverse Financial Assets
• Spiked sample covariance matrices with possibly multiple bulk components
• High-dimensional GARCH process segmentation with an application to Value-at-Risk
• Aspects of a phase transition in high-dimensional random geometry
• II. High Dimensional Estimation under Weak Moment Assumptions: Structured Recovery and Matrix Estimation
• Numerical Implementation of the QuEST Function
• Optimal trading strategies - a time series approach
• Near optimal sample complexity for matrix and tensor normal models via geodesic convexity
• Second Degree Model for Multi-Compression and Recovery of Distributed Signals
• Eigenvector overlaps in large sample covariance matrices and nonlinear shrinkage estimators
• A Fourier analytique approach to Gaussian mixture learning
• Maximum Likelihood Estimation for Brownian Motion Tree Models Based on One Sample