Abstract:This paper considers the problem of testing the equality of two unspecified distributions. The classical omnibus tests such as the Kolmogorov-Smirnov and Cramèr-von Mises are known to suffer from low power against essentially all but location-scale alternatives. We propose a new two-sample test that modifies the Neyman's smooth test and extend it to the multivariate case based on the idea of projection pursue. The asymptotic null property of the test and its power against local alternatives are studied. The multiplier bootstrap method is employed to compute the critical value of the multivariate test. We establish validity of the bootstrap approximation in the case where the dimension is allowed to grow with the sample size. Numerical studies show that the new testing procedures perform well even for small sample sizes and are powerful in detecting local features or high-frequency components.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is: given two sets of samples, how to effectively test whether these two samples are from the same distribution. Specifically, the authors are concerned with the problem of low power in traditional omnibus test methods (such as the Kolmogorov - Smirnov and Cramér - von Mises tests) when detecting alternative hypotheses other than location - scale. To solve this problem, the paper proposes a new two - sample smooth test method, which is based on Neyman's smooth test principle and extended to the multivariate case. Through the idea of projection pursuit, this method can effectively detect local features or high - frequency components in high - dimensional data. ### Specific Problem Description 1. **Limitations of Traditional Methods**: - Traditional omnibus test methods (such as the Kolmogorov - Smirnov and Cramér - von Mises tests) have low power when detecting cases other than location - scale alternative hypotheses. - These methods perform poorly when detecting density functions containing high - frequency components or local features. 2. **Goals of the New Method**: - Propose a new two - sample test method to improve the detection ability for various alternative hypotheses. - This method should be able to handle high - dimensional data and also show good performance in the case of small samples. ### Innovations of the New Method 1. **Improvement Based on Neyman's Smooth Test**: - A new two - sample smooth test method is proposed, which modifies Neyman's smooth test and is extended to the multivariate case. - Through the idea of projection pursuit, the multivariate problem is transformed into a series of one - dimensional problems for processing. 2. **Theoretical Analysis and Numerical Verification**: - The asymptotic properties of the new test method under the null hypothesis and its detection ability for local alternative hypotheses are studied. - The multiplier bootstrap method is used to calculate the critical value to ensure the effectiveness of the method in practical applications. - Through numerical research, it is verified that the new method can also show good performance in the case of small samples. ### Summary This paper aims to solve the problem of low power in traditional two - sample distribution equality test methods when detecting complex alternative hypotheses. By introducing a new method based on Neyman's smooth test principle and combining projection pursuit technology, the authors provide a more powerful tool to detect whether two samples are from the same distribution, especially in the case of high - dimensional data and small samples.

Two-Sample Smooth Tests for the Equality of Distributions

Tests for the Equality of Two Distributions Based on Kolmgorov-Smirnov Statistic

Two-Sample Test for Sparse High Dimensional Multinomial Distributions

The Asymptotic Distribution and Berry–Esseen Bound of a New Test for Independence in High Dimension with an Application to Stochastic Optimization

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

One and two sample Dvoretzky-Kiefer-Wolfowitz-Massart type inequalities for differing underlying distributions

Generalized kernel two-sample tests

The AUGUST Two-Sample Test: Powerful, Interpretable, and Fast

A Kernel Method for the Two-Sample Problem

Two sample test for covariance matrices in ultra-high dimension

A distance based two-sample test of means difference for multivariate datasets

A uniform kernel trick for high-dimensional two-sample problems

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

Double verification for two‐sample covariance matrices test

Testing distributional equality for functional random variables

Smooth test for equality of copulas

Bayesian Optimal Two-sample Tests in High-dimension

A class of nonparametric tests for the two-sample problem based on order statistics

A bootstrap test for equality of variances

Two Sample Tests for High Dimensional Covariance Matrices

The power of visualizing distributional differences: formal graphical n-sample tests