A p-value based dimensionality reduction test for high dimensional means

With the rapid development of modern computing techniques, high-dimensional data are increasingly encountered in many studies. In this paper, we propose a three-step method to study the mean testing problem. The proposed test is based on the p-values calculated from the univariate tests and the dime...

Full description

Saved in:
Bibliographic Details
Published inStatistics (Berlin, DDR) Vol. 57; no. 2; pp. 282 - 299
Main Authors Fang, Hongyan, Yao, Chunyu, Yang, Wenzhi, Wang, Xuejun, Xu, Huang
Format Journal Article
LanguageEnglish
Published Abingdon Taylor & Francis 04.03.2023
Taylor & Francis Ltd
Subjects
Computation
computation efficiency
Gene expression
hotelling
Hypothesis testing high-dimensional data
mean vector
multivariate normal
Normal distribution
Reduction
Simulation
test
Vectors
Online AccessGet full text

Cover

More Information
Summary:With the rapid development of modern computing techniques, high-dimensional data are increasingly encountered in many studies. In this paper, we propose a three-step method to study the mean testing problem. The proposed test is based on the p-values calculated from the univariate tests and the dimension reduction method. Since it does not require explicit conditions of data dimension and sample size, we can use it to solve the mean testing problem of high-dimensional data, especially when the data dimension is much larger than the sample size. The new method can be implemented for the normal and non-normal distribution, which has a wide application. Various simulations are conducted to compare the testing power of the new method and the existing tests. The comparison shows that the new method has higher testing power. We also apply the proposed method to a real example of gene expression data.
ISSN:0233-1888
1029-4910
DOI:10.1080/02331888.2023.2179627