A Class of Quasi-Distribution-free Multivariate Multisample Lepage Tests and Their Applications in High-Dimensional Data Study  
Author Kazuki Matsuo

 

Co-Author(s) Amitava Mukherjee;  Hidetoshi Murakami

 

Abstract In the present paper, we investigate some MVMS-Lepage tests based on PCA and ICA, namely, MVMS-PCS-Lepage and MVMS-ICA-Lepage tests. We compare their performance with traditional MVMS-ORD-Lepage, MVMS-SMD-Lepage, and various MVMS-KW tests. Proposed tests are highly efficient in detecting the difference in the location vector, scale matrix or both irrespective of the sample sizes from various populations, which can be any natural number, not necessarily all equal. Monte-Carlo experiments are carried out using various hypothetical population distributions, and the proposed tests are established as quasidistribution free. Computation of permutation-based p-values in the presence of ties is also discussed. We offer an illustration of the proposed tests using real data. The paper concludes with some remarks and directions for future research.

 

Keywords high-dimensional tests; independent component analysis; Lepage test; Multisample comparisons; principal component analysis
   
    Article #:  DSBFI23-5
 
Proceedings of 2nd ISSAT International Conference on Data Science in Business, Finance and Industry
January 8-10, 2023 - Da Nang, Vietnam