期刊名称：International Journal of Academic Research in Business and Social Sciences
电子版ISSN：2222-6990
出版年度：2018
卷号：8
期号：12
页码：1773-1784
DOI：10.6007/IJARBSS/v8-i12/5322
语种：English
出版社：Human Resource Management Academic Research Society
摘要：The existing procedures for structural equation modelling involve in goodness of fit test for the sample covariance matrix by the structural model can no longer work in high dimensional datasets. The sample covariance can easily be influenced by outlier’s presence in the datasets. This affect the estimation of the sample mean and sample covariance not being accurate and as well cause inefficiency in the computation. Therefore, there is need to suggest a robust covariance estimator the L_1-median with Weiszfeld algorithm that will resolve the outliers problem in high dimensional dataset. This test is subjected to conditions of sample size (n), variables (p) percentages of outliers (ε) with α=0.05. Simulation study carried out and the results show that when variable is minor both test performed but the new robust covariance test is better. At both middle and greater variables the existing test cannot compute the rate when p>n cases. Generally, the results shows that the newly incorporated robust covariance test performed better compare to the existing test.
