Testing simultaneously different covariance block diagonal structures - the multi-sample case

Area of Science:

• Multivariate statistics
• Statistical hypothesis testing
• Covariance matrix analysis

Background:

• Testing equality of several covariance matrices is a common problem in multivariate analysis.
• Existing methods may not efficiently handle complex structural hypotheses, such as block diagonality.

Purpose of the Study:

• To develop a likelihood ratio test for simultaneously testing matrix equality and specific block diagonal structures.
• To derive the null distribution and moments of the likelihood ratio statistic.
• To provide accurate approximations for practical application.

Main Methods:

• Development of a novel likelihood ratio test statistic.
• Derivation of the theoretical distribution and h-th null moment of the statistic.
• Construction of near-exact approximations for the test statistic.

Main Results:

• A likelihood ratio test is established for simultaneous equality and block diagonal structure testing.
• The h-th null moment of the statistic is derived, aiding distributional analysis.
• Near-exact approximations are developed, enhancing practical utility.

Conclusions:

• The developed likelihood ratio test is effective for complex covariance matrix structures.
• The near-exact approximations improve the test's applicability to real-world data.
• Numerical studies and simulations validate the test's performance and approximation accuracy.