Tests for high dimensional data based on means, spatial signs and spatial ranks

TL;DR

This paper compares mean-based and spatial sign/rank-based tests for high-dimensional data, showing their asymptotic power equivalence under certain conditions and potential superiority of spatial methods with strong dependence, with practical implications for test accuracy.

Contribution

It provides new asymptotic results for high-dimensional tests based on means, spatial signs, and ranks, highlighting conditions under which spatial methods outperform mean-based tests.

Findings

Power of mean and spatial sign/rank tests are asymptotically equivalent under certain conditions.

Spatial sign and rank tests can be more powerful with strong dependence among variables.

Asymptotic approximations for spatial sign/rank tests are accurate on simulated and real data.

Abstract

Tests based on sample mean vectors and sample spatial signs have been studied in the recent literature for high dimensional data with the dimension larger than the sample size. For suitable sequences of alternatives, we show that the powers of the mean based tests and the tests based on spatial signs and ranks tend to be same as the data dimension grows to infinity for any sample size, when the coordinate variables satisfy appropriate mixing conditions. Further, their limiting powers do not depend on the heaviness of the tails of the distributions. This is in striking contrast to the asymptotic results obtained in the classical multivariate setup. On the other hand, we show that in the presence of stronger dependence among the coordinate variables, the spatial sign and rank based tests for high dimensional data can be asymptotically more powerful than the mean based tests if in addition to the data dimension, the sample size also grows to infinity. The sizes of some mean based tests for high dimensional data studied in the recent literature are observed to be significantly different from their nominal levels. This is due to the inadequacy of the asymptotic approximations used for the distributions of those test statistics. However, our asymptotic approximations for the tests based on spatial signs and ranks are observed to work well when the tests are applied on a variety of simulated and real datasets.