Comparing a large number of multivariate distributions
From MaRDI portal
Publication:2214253
Abstract: In this paper, we propose a test for the equality of multiple distributions based on kernel mean embeddings. Our framework provides a flexible way to handle multivariate or even high-dimensional data by virtue of kernel methods and allows the number of distributions to increase with the sample size. This is in contrast to previous studies that have been mostly restricted to classical low-dimensional settings with a fixed number of distributions. By building on Cramer-type moderate deviation for degenerate two-sample V-statistics, we derive the limiting null distribution of the test statistic and show that it converges to a Gumbel distribution. The limiting distribution, however, depends on an infinite number of nuisance parameters, which makes it infeasible for use in practice. To address this issue, the proposed test is implemented via the permutation procedure and is shown to be minimax rate optimal against sparse alternatives. During our analysis, an exponential concentration inequality for the permuted test statistic is developed which may be of independent interest.
The author introduces a new nonparametric \(K\)-sample test based on the maximum mean discrepancy. The limiting distribution of the proposed test statistic is derived based on the Cramér-type moderate deviation for degenerate two-sample \(V\)-statistics. However, the obtained limiting distribution relies on an infinite number of nuisance parameters, which are intractable in general. Therefore, a permutation approach to determine the cut-off value of the test is used. A concentration inequality for the proposed test statistic with a sharp exponential tail bound under permutations is presented. This result is used to study the power of the suggested permutation test when both \(K\) and \(N\) are large. The author claims that his vast simulations show that the proposed permutation test is powerful against sparse alternatives. The proofs are not presented and can be found in \url{doi:10.3150/20-BEJ1244SUPP}.
Recommendations
- Multiple comparisons for a large number of parameters
- Comparison of multivariate distributions using quantile-quantile plots and related tests
- scientific article; zbMATH DE number 4117659
- Multiple comparisons of mean vectors with large dimension under general conditions
- On multivariate tests for comparing means
- scientific article; zbMATH DE number 3864304
- Stochastic comparison of multivariate random sums
- Comparing distributions by multiple testing across quantiles or CDF values
- scientific article; zbMATH DE number 4001245
Cites work
- \(K\)-sample problem using strong approximations of empirical copula processes
- \(k\)-sample test based on the common area of kernel density estimators
- A kernel two-sample test
- A nonparametric approach to high-dimensional k-sample comparison problems
- Asymptotic Theory of Certain "Goodness of Fit" Criteria Based on Stochastic Processes
- Circular law for random matrices with exchangeable entries
- Comparing distributions
- Concentration inequalities for randomly permuted sums
- Concentration inequalities. A nonasymptotic theory of independence
- Concentration of normalized sums and a central limit theorem for noncorrelated random variables
- Cramér-von Mises and characteristic function tests for the two and \(k\)-sample problems with dependent data
- DISCO analysis: A nonparametric extension of analysis of variance
- Distribution-free tests of independence in high dimensions
- Equivalence of distance-based and RKHS-based statistics in hypothesis testing
- Global testing under sparse alternatives: ANOVA, multiple comparisons and the higher criticism
- High-dimensional probability. An introduction with applications in data science
- High-dimensional sparse MANOVA
- scientific article; zbMATH DE number 4170917 (Why is no real title available?)
- K-Sample Analogues of the Kolmogorov-Smirnov and Cramer-V. Mises Tests
- Kernel Mean Embedding of Distributions: A Review and Beyond
- Minimax estimation of kernel mean embeddings
- Nonparametric K-sample tests via dynamic slicing
- On a new multivariate two-sample test.
- On convergence rates of suprema
- On the asymptotic power of some k-sample statistics based on the multivariate empirical process
- Optimal sparse segment identification with application in copy number variation analysis
- Permutation p-value approximation via generalized Stolarsky invariance
- Power enhancement in high-dimensional cross-sectional tests
- Several k-Sample Kolmogorov-Smirnov Tests
- Sign-based test for mean vector in high-dimensional and sparse settings
- Stein's method for concentration inequalities
- Testing equality of a large number of densities
- Testing Statistical Hypotheses
- Tests for the multivariatek-sample problem based on the empirical characteristic function
- The tight constant in the Dvoretzky-Kiefer-Wolfowitz inequality
- Two moments suffice for Poisson approximations: The Chen-Stein method
- Two-Sample Covariance Matrix Testing and Support Recovery in High-Dimensional and Sparse Settings
- Two-Sample Test of High Dimensional Means Under Dependence
- Universality, Characteristic Kernels and RKHS Embedding of Measures
Cited in
(16)- Local permutation tests for conditional independence
- Minimax optimality of permutation tests
- Testing the equality of a large number of populations
- Testing equality of a large number of densities
- scientific article; zbMATH DE number 3864304 (Why is no real title available?)
- Comparison of multivariate distributions using quantile-quantile plots and related tests
- scientific article; zbMATH DE number 895060 (Why is no real title available?)
- Graphical Comparison of High‐Dimensional Distributions
- Homogeneity of marginal distributions for a large number of populations
- Minimax nonparametric multi-sample test under smoothing
- A conditional distribution function-based measure for independence and K-sample tests in multivariate data
- The \(\mathrm{k}\)-sample problem using gini covariance for large \(\mathrm{k}\)
- k-sample inference via multimarginal optimal transport
- A review of multivariate permutation tests: findings and trends
- Graph-based multisample comparison with application to feature selection for multi-category responses
- Testing the equality of estimable parameters across many populations
This page was built for publication: Comparing a large number of multivariate distributions
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2214253)
