报告时间：11月25日周五10:00
报告地点：统计楼105
报告题目：Testing for block correlation of several sets of high dimensional random variables
报告人：郑术蓉教授 东北师大
报告简介：
This paper intends to develop a unified test on block correlation of high dimensional random vectors. This test is not only valid for low dimension or high dimension, but also is valid for Gaussian population or non-Gaussian population. Simulation studies show that Type I errors can be well kept, that is, the limiting null distributioncan well approximate to the true null distribution of the proposed statistic. Moreover, under the alternative hypothesis, the limiting distribution is also derived and the asymptotic theoretical power function is given. When the dimension is smaller than the sample size, simulation studies are conducted to compare our proposed test with the existing tests for Gaussian population. For comparison of empirical powers, our proposed test outperforms all the existing tests. Even when the population is non-Gaussian and the dimension is more greater than the sample size, our proposed test performs well for empirical powers or empirical Type I error rates.