Testing Several Correlation Matrices Using Robust Approach

• Main Authors: Atiany, Tareq A.M., Sharif, Shamshuritawati
• Format: Article
• Language: English
• Published: Asian Network for Scientific Information 2017
• Subjects: Q Science (General)
• Online Access: http://repo.uum.edu.my/24427/1/AJSR%2011%201%202018%2084-95.pdf; http://repo.uum.edu.my/24427/; http://doi.org/10.3923/ajsr.2018.84.95

Background and Objective: The performance of classical Jennrich (J) statistic using classical estimators suffers from masking effects. To relieve the problem, robust estimators are recommended. In this study, a robust Jennrich statistic was proposed based on a S estimator (JS) and M estimator (JM) as alternative to the J statistic, which has good properties. Methodology: In the simulation study, the performance of proposed test is assessed in terms of a type I error and the power of test. The performance comparison between classical J, JS and JM statistics are conducted under several conditions. Results: The results of simulation study showed that JS statistic has a competitive performance comparative to a JM statistic and the J statistic. Conclusion: It was concluded that JS statistic is robust for testing the equality of two or more difference correlation matrices when the data contains outlier. © 2018 Shamshuritawati Sharif and Tareq Ahmed Mahmoud Atiany.