TEST FOR HIGH DIMENSIONAL CORRELATION MATRICES.

Area of Science:

• Statistics
• High-Dimensional Data Analysis
• Multivariate Analysis

Background:

• Testing correlation structures is crucial in many applications but presents theoretical challenges.
• Existing methods often struggle with high-dimensional data where sample size and dimension grow concurrently.

Purpose of the Study:

• To develop a unified framework for testing correlation structures in one-, two-, and multiple-sample problems.
• To address challenges in high-dimensional settings where both sample size and data dimension tend to infinity.
• To create test statistics robust to both dense and sparse alternatives.

Main Methods:

• Development of novel test statistics for high-dimensional correlation structure testing.
• Systematic investigation of asymptotic null distribution, power function, and unbiasedness.
• Theoretical analysis addressing the non-independency of sample correlation matrices.
• Empirical validation through simulation studies and real-world data analysis.

Main Results:

• The proposed framework provides versatile test statistics for high-dimensional correlation testing.
• The test statistics are designed to be effective for both dense and sparse alternatives.
• Asymptotic properties (null distribution, power, unbiasedness) of the test statistics are rigorously analyzed.
• Simulation and real data analyses confirm the practicability and broad applicability of the methods.

Conclusions:

• The developed general framework offers a robust solution for testing correlation structures in high-dimensional data.
• The proposed test statistics demonstrate effectiveness and versatility across different sample testing scenarios.
• The study contributes significant theoretical insights into high-dimensional statistics and correlation analysis.