Finite Sample t-Tests for High-Dimensional Means

TL;DR

This paper develops finite-sample t-tests for high-dimensional mean vectors that remain accurate with very small, fixed sample sizes, addressing size distortion issues in traditional asymptotic tests.

Contribution

It establishes asymptotic t-distributions for U-statistics in high-dimensional, small-sample settings, enabling more reliable testing.

Findings

Proposed tests maintain accurate sizes across various dimensions and sample sizes.

Simulation studies validate the theoretical asymptotic distributions.

Application to fMRI data demonstrates practical utility.

Abstract

Size distortion can occur if an asymptotic testing procedure requiring diverging sample sizes, is implemented to data with very small sample sizes. In this paper, we consider one-sample and two-sample tests for mean vectors when data are high-dimensional but sample sizes are very small. We establish asymptotic t-distributions of one-sample and two-sample U-statistics, which only require data dimensionality to diverge but sample sizes to be fixed and no less than 3. Simulation studies confirm the theoretical results that the proposed tests maintain accurate empirical sizes for a wide range of sample sizes and data dimensionalities. We apply the proposed tests to an fMRI dataset to demonstrate the practical implementation of the methods.