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.

On some characterizations and multidimensional criteria for testing homogeneity, symmetry and independence

Tests for a Multiple-Sample Problem in High Dimensions

A Robust and Nonparametric Two-Sample Test in High Dimensions

On high-dimensional modifications of some graph-based two-sample tests

Multivariate Symmetry: Distribution-Free Testing via Optimal Transport

Paired-sample Tests for Homogeneity With/without Confounding Variables

Conditional symmetry test based on empirical characteristic function

Multivariate two-sample test statistics based on data depth

Testing Homogeneity in High Dimensional Data Through Random Projections

Reducing multivariate independence testing to two bivariate means comparisons

Two-Sample Test for Sparse High Dimensional Multinomial Distributions

New graph-based multi-sample tests for high-dimensional and non-Euclidean data

Multivariate Rank-Based Distribution-Free Nonparametric Testing Using Measure Transportation

Homogeneity Test of Multi-Sample Covariance Matrices in High Dimensions

A Hyperbolic Divergence Based Nonparametric Test for Two‐sample Multivariate Distributions

Symmetrical Independence Tests for Two Random Vectors with Arbitrary Dimensional Graphs

Distribution-free tests of multivariate independence based on center-outward quadrant, Spearman, Kendall, and van der Waerden statistics

Two-sample tests for multivariate functional data with applications

RANK-BASED INDICES FOR TESTING INDEPENDENCE BETWEEN TWO HIGH-DIMENSIONAL VECTORS

Log-Rank-Type Tests for Equality of Distributions in High-Dimensional Spaces.

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