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

DOI10.1214/19-AOS1934MaRDI QIDQ1996774zbMATH OpenFDO

Authors Changbo Zhu, Xianyang Zhang, Shun Yao, Xiaofeng Shao

Publication date 26 February 2021

Published in The Annals of Statistics (Search for Journal in Brave)

Full work available at URL https://arxiv.org/abs/1902.03291, https://projecteuclid.org/euclid.aos/1607677239

zbMATH Keywords

\(\mathcal{U}\)-statisticsdistance covariancehigh dimensionalityHilbert-Schmidt independence criterionindependence testreproducing kernel Hilbert space (RKHS)

Mathematics Subject Classification ID

Hilbert spaces with reproducing kernels (= (proper) functional Hilbert spaces, including de Branges-Rovnyak and other structured spaces) (46E22) Interacting random processes; statistical mechanics type models; percolation theory (60K35) Nonparametric hypothesis testing (62G10) Asymptotic properties of nonparametric inference (62G20) Applications of statistics to environmental and related topics (62P12) Seismology (including tsunami modeling), earthquakes (86A15)

Abstract: In this paper, we study distance covariance, Hilbert-Schmidt covariance (aka Hilbert-Schmidt independence criterion [Gretton et al. (2008)]) 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 distance/Hilbert-Schmidt covariance based test can only capture linear dependence in high dimension. As a consequence, the distance correlation based t-test developed by Szekely and Rizzo (2013) 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 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.

Review text

The authors consider conditions under which \[\mathrm{d}\mathrm{Cov}_n^2\left(\mathbf{X},\mathbf{Y}\right)\approx\frac{1}{\tau}\sum_{i=1}^p\sum_{j=1}^q\mathrm{cov}_n^2\left(\mathcal{X}_i,\mathcal{Y}_j\right).\] Here \(\mathrm{d}\mathrm{Cov}_n^2\left(\mathbf{X},\mathbf{Y}\right)\) denotes the umbiased sample distance covariance, \[ \mathbf{X}=\left(X_1,X_2,\ldots,X_n\right)^\intercal =\left(\mathcal{X}_1,\mathcal{X}_2,\ldots,\mathcal{X}_p\right), \] \[ \mathbf{Y}=\left(Y_1,Y_2,\ldots,Y_n\right)^\intercal=\left(\mathcal{Y}_1,\mathcal{Y}_2,\ldots,\mathcal{Y}_q\right) \] are the sample matrices, where \(X_k\mathop{=}\limits^{d}X\) and \(Y_k\mathop{=}\limits^{d}Y\) are independent samples of two random vectors \(X=(x_1,x_2,\ldots,x_p)\in\mathbb{R}^p\) and \(Y=(y_1,y_2,\ldots,y_q)\in\mathbb{R}^q\) with finite componentwise second moments, and \(\mathcal{X}_i\), \(\mathcal{Y}_j\) are the componentwise samples. In addition, \(\tau\) denotes a quantity depending on the marginal distributions of \(X\) and \(Y\) as well as \(p\) and \(q\), and \(\mathrm{cov}_n\left(\mathcal{X}_i,\mathcal{Y}_j\right)\) is an unbiased sample estimate of \(\mathrm{cov}(x_i,y_j)\). The above approximate equality is considered as \(p,q\) tend to infinity, and \(n\) can either be fixed or grows to infinity at a slower rate. The paper is the first work on the connection between sample distance covariance and sample covariance.

Recommendations

• On distance covariance in metric and Hilbert spaces
• A basic treatment of the distance covariance
• Detecting independence of random vectors: generalized distance covariance and Gaussian covariance
• Measuring and testing dependence by correlation of distances
• Brownian distance covariance

Cites work

• A consistent test of independence based on a sign covariance related to Kendall's tau
• A kernel two-sample test
• A Non-Parametric Test of Independence
• A Nonparametric Test of Independence Between Two Vectors
• A survey of high dimension low sample size asymptotics
• An asymptotic decomposition for multivariate distribution-free tests of independence
• Asymptotic theory of statistics and probability
• Brownian distance covariance
• Conditional mean and quantile dependence testing in high dimension
• Cramer-von Mises tests for independence
• Distance covariance in metric spaces
• Distance metrics for measuring joint dependence with application to causal inference
• Distribution Free Tests of Independence Based on the Sample Distribution Function
• Equivalence of distance-based and RKHS-based statistics in hypothesis testing
• Equivalence of kernel machine regression and kernel distance covariance for multidimensional phenotype association studies
• Feature screening via distance correlation learning
• Geometric Representation of High Dimension, Low Sample Size Data
• scientific article; zbMATH DE number 509167 (Why is no real title available?)
• Independence test for high dimensional data based on regularized canonical correlation coefficients
• Introduction to strong mixing conditions. Vol. 1.
• Martingale difference correlation and its use in high-dimensional variable screening
• Measuring and testing dependence by correlation of distances
• Measuring nonlinear dependence in time-series, a distance correlation approach
• Multivariate nonparametric tests for independence
• Nonlinear system theory: Another look at dependence
• Nonparametric independence testing via mutual information
• Partial distance correlation with methods for dissimilarities
• Partial martingale difference correlation
• PCA consistency in high dimension, low sample size context
• Projection correlation between two random vectors
• Sign test of independence between two random vectors.
• Testing for independence by the empirical characteristic function
• Testing independence among a large number of high-dimensional random vectors
• Testing independence in high dimensions with sums of rank correlations
• Testing mutual independence in high dimension via distance covariance
• The affinely invariant distance correlation
• The distance correlation \(t\)-test of independence in high dimension
• The high-dimension, low-sample-size geometric representation holds under mild conditions
• The notion of \(\psi \)-weak dependence and its applications to bootstrapping time series

Cited in

(40)

• Interpoint distance based two sample tests in high dimension
• Asymptotic distributions of high-dimensional distance correlation inference
• A new framework for distance and kernel-based metrics in high dimensions
• Adaptive test of independence based on HSIC measures
• Distance covariance for random fields
• Testing for independence of high-dimensional variables: V-coefficient based approach
• Independence tests with random subspace of two random vectors in high dimension
• Testing mutual independence in high dimension via distance covariance
• On distance covariance in metric and Hilbert spaces
• Marginal Distance and Hilbert-Schmidt Covariances-Based Independence Tests for Multivariate Functional Data
• Eigenvalue Distribution of a High-Dimensional Distance Covariance Matrix With Application
• Generalization of the HSIC and distance covariance using PDI kernels
• Asymptotic normality of Gini correlation in high dimension with applications to the \(K\)-sample problem
• Rank-based indices for testing independence between two high-dimensional vectors
• Statistical Inferences for Complex Dependence of Multimodal Imaging Data
• Application of distance standard deviation in functional data analysis
• Asymptotic Distribution-Free Independence Test for High-Dimension Data
• Generalized kernel distance covariance in high dimensions: non-null CLTs and power universality
• A conditional distribution function-based measure for independence and K-sample tests in multivariate data
• A class of robust independence tests based on weighted integrals of empirical characteristic functions
• A slicing-free perspective to sufficient dimension reduction: selective review and recent developments
• Testing for independence in high dimensions based on characteristic covariance
• Distance correlation in the presence of measurement errors
• A robust nonparametric test for conditional symmetry in high dimension
• High-Dimensional Block Diagonal Covariance Structure Detection Using Singular Vectors
• Measuring dependence between functional data via projection Hilbert-Schmidt covariance
• A distance covariance test of independence in high dimension, low sample size contexts
• Berry-Esseen bounds for degenerate U-statistics with application to distance correlation
• An efficient and distribution-free symmetry test for high-dimensional data based on energy statistics and random projections
• Multivariate differential association analysis
• Kernel Angle Dependence Measures in Metric Spaces
• A kernel independence test using projection-based measure in high-dimension
• Testing independence for sparse longitudinal data
• Change point detection in high-dimensional data with U-statistics
• Generalized Spectral Tests for Multivariate Martingale Difference Hypotheses
• Detect complete dependence via trace correlation in the presence of matrix-valued random objects
• Testing independence between high-dimensional random vectors using rank-based max-sum tests
• Test of independence using generalized distance correlation
• Distribution-free robust independence test for high-dimensional data via semi-Grothendieck's covariance
• Robust independence test of functional variables via Gupta angle covariance

This page was built for publication: Distance-based and RKHS-based dependence metrics in high dimension

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q1996774)