Abstract:In this paper, we study distance covariance, Hilbert–Schmidt covariance (aka Hilbert–Schmidt independence criterion [In Advances in Neural Information Processing Systems (2008) 585–592]) and related independence tests under the high dimensional scenario. We show that the sample distance/Hilbert–Schmidt covariance between two random vectors can be approximated by the sum of squared componentwise sample cross-covariances up to an asymptotically constant factor, which indicates that the standard distance/Hilbert–Schmidt covariance based test can only capture linear dependence in high dimension. Under the assumption that the components within each high dimensional vector are weakly dependent, the distance correlation based $t$ test developed by Székely and Rizzo (J. Multivariate Anal. 117 (2013) 193–213) for independence is shown to have trivial limiting power when the two random vectors are nonlinearly dependent but component-wisely uncorrelated. This new and surprising phenomenon, which seems to be discovered and carefully studied for the first time, is further confirmed in our simulation study. As a remedy, we propose tests based on an aggregation of marginal sample distance/Hilbert–Schmidt covariances and show their superior power behavior against their joint counterparts in simulations. We further extend the distance correlation based $t$ test to those based on Hilbert–Schmidt covariance and marginal distance/Hilbert–Schmidt covariance. A novel unified approach is developed to analyze the studentized sample distance/Hilbert–Schmidt covariance as well as the studentized sample marginal distance covariance under both null and alternative hypothesis. Our theoretical and simulation results shed light on the limitation of distance/Hilbert–Schmidt covariance when used jointly in the high dimensional setting and suggest the aggregation of marginal distance/Hilbert–Schmidt covariance as a useful alternative.

Detecting Manifold Dependences of Multivariate Data with Total Correlation.

Relative Synergy Coefficient: A Novel Way to Detect Variable Interaction in Large Dataset

Detecting and Visualizing Profile Correlation in Subspace

Unbiased Multivariate Correlation Analysis.

Measures of Correlation for Multiple Variables

Discovering and Deciphering Relationships Across Disparate Data Modalities

Measures of conditional dependence for nonlinearity, asymmetry and beyond

A Riemannian covariance for manifold-valued data

Automated Fractional Limb Volume Measurements Improve the Precision of Birth Weight Predictions in Late Third‐Trimester Fetuses

Azadkia-Chatterjee's correlation coefficient adapts to manifold data

Multivariate dependence and genetic networks inference

On eigenvalues of a high-dimensional Kendall's rank correlation matrix with dependence

Estimating Total Correlation with Mutual Information Estimators

Mining maximal correlated member clusters in high dimensional database

Ordinal pattern dependence as a multivariate dependence measure

Intrinsic Dimension Correlation: uncovering nonlinear connections in multimodal representations

A unified framework for correlation mining in ultra-high dimension

A New Class of Nonsymmetric Multivariate Dependence Measures

High-dimensional sparse MANOVA

Distance-based and RKHS-based dependence metrics in high dimension

Quantifying Multivariate Graph Dependencies: Theory and Estimation for Multiplex Graphs