Dimension-agnostic inference using cross U-statistics

DOI10.3150/23-BEJ1613MaRDI QIDQ6178581zbMATH OpenOpenAlexFDO

Authors Ilmun Kim, Aaditya Ramdas

Publication date 16 January 2024

Published in Bernoulli (Search for Journal in Brave)

Full work available at URL https://arxiv.org/abs/2011.05068

zbMATH Keywords

minimax optimalitysample splittingstudentizationdegenerate U-statisticshigh-dimensional limits

Mathematics Subject Classification ID

Nonparametric inference (62Gxx) Parametric inference (62Fxx) Multivariate analysis (62Hxx)

Abstract: Classical asymptotic theory for statistical inference usually involves calibrating a statistic by fixing the dimension $d$ while letting the sample size $n$ increase to infinity. Recently, much effort has been dedicated towards understanding how these methods behave in high-dimensional settings, where $d$ and $n$ both increase to infinity together. This often leads to different inference procedures, depending on the assumptions about the dimensionality, leaving the practitioner in a bind: given a dataset with 100 samples in 20 dimensions, should they calibrate by assuming $nggd$, or $d/napprox0.2$? This paper considers the goal of dimension-agnostic inference; developing methods whose validity does not depend on any assumption on $d$ versus $n$. We introduce an approach that uses variational representations of existing test statistics along with sample splitting and self-normalization to produce a refined test statistic with a Gaussian limiting distribution, regardless of how $d$ scales with $n$. The resulting statistic can be viewed as a careful modification of degenerate U-statistics, dropping diagonal blocks and retaining off-diagonal blocks. We exemplify our technique for some classical problems including one-sample mean and covariance testing, and show that our tests have minimax rate-optimal power against appropriate local alternatives. In most settings, our cross U-statistic matches the high-dimensional power of the corresponding (degenerate) U-statistic up to a $sqrt2$ factor.

Cites work

• scientific article; zbMATH DE number 3551712 (Why is no real title available?)
• scientific article; zbMATH DE number 1964693 (Why is no real title available?)
• scientific article; zbMATH DE number 889593 (Why is no real title available?)
• A feasible high dimensional randomization test for the mean vector
• A kernel two-sample test
• A modern maximum-likelihood theory for high-dimensional logistic regression
• A new test for multivariate normality
• A note on data-splitting for the evaluation of significance levels
• A note on testing the covariance matrix for large dimension
• A one-sample test for normality with kernel methods
• A review of 20 years of naive tests of significance for high-dimensional mean vectors and covariance matrices
• A test for the mean vector with fewer observations than the dimension
• A two-sample test for high-dimensional data with applications to gene-set testing
• Asymptotic behavior of M estimators of p regression parameters when \(p^ 2/n\) is large. II: Normal approximation
• Asymptotic behavior of M-estimators of p regression parameters when p^ 2/n is large. I. Consistency
• Asymptotic normality of a consistent estimator of maximum mean discrepancy in Hilbert space
• Asymptotic normality of quadratic estimators
• Bootstrapping and sample splitting for high-dimensional, assumption-lean inference
• Can we trust the bootstrap in high-dimensions? The case of linear models
• Central limit theorem for integrated square error of multivariate nonparametric density estimators
• Central limit theorems and bootstrap in high dimensions
• Central limit theorems for classical likelihood ratio tests for high-dimensional normal distributions
• Classification accuracy as a proxy for two-sample testing
• Conditional Distance Correlation
• Conditional mean and quantile dependence testing in high dimension
• Distribution and quantile functions, ranks and signs in dimension \(d\): a measure transportation approach
• Distribution-Free Consistent Independence Tests via Center-Outward Ranks and Signs
• Distribution-free predictive inference for regression
• Estimation and Inference of Heterogeneous Treatment Effects using Random Forests
• Estimation of integrated squared density derivatives
• Exact and Approximate Stepdown Methods for Multiple Hypothesis Testing
• Exact bounds on the closeness between the Student and standard normal distributions
• Gaussian universal likelihood ratio testing
• Goodness-of-fit Testing in High Dimensional Generalized Linear Models
• High-dimensional probability. An introduction with applications in data science
• Interaction screening for ultrahigh-dimensional data
• Minimax Euclidean separation rates for testing convex hypotheses in \(\mathbb{R}^{d}\)
• Minimax optimality of permutation tests
• Modification of some goodness-of-fit statistics to yield asymptotically normal null distributions
• Monge-Kantorovich depth, quantiles, ranks and signs
• Multinomial goodness-of-fit based on \(U\)-statistics: high-dimensional asymptotic and minimax optimality
• Multivariate Rank-Based Distribution-Free Nonparametric Testing Using Measure Transportation
• Non-asymptotic minimax rates of testing in signal detection
• Nonparametric goodness-of-fit testing under Gaussian models
• On some test criteria for covariance matrix
• On the power of conditional independence testing under model-X
• Optimal hypothesis testing for high dimensional covariance matrices
• Quantifying uncertainty in random forests via confidence intervals and hypothesis tests
• Robust multivariate nonparametric tests via projection averaging
• Some properties of incomplete U-statistics
• Split sample methods for constructing confidence intervals for binomial and Poisson parameters
• Student's t-Test Under Symmetry Conditions
• Testing Statistical Hypotheses
• Tests for high-dimensional covariance matrices
• Tests for high-dimensional regression coefficients with factorial designs
• The Berry-Esseen bound for Student's statistic
• The Holdout Randomization Test for Feature Selection in Black Box Models
• The Large-Sample Distribution of the Likelihood Ratio for Testing Composite Hypotheses
• Two-Sample Test of High Dimensional Means Under Dependence
• Universal inference

Cited in

(1)

• Limit theorems for a class of processes generalizing the U -empirical process

This page was built for publication: Dimension-agnostic inference using cross U-statistics

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6178581)