Two Sample Mean Test Of High Dimensional Data Based On Bootstrap Method

With the rapid development of society and science and technology,high dimen-sional data have emerged in many fields,such as gene expression value and single nucleotide polymorphism in microarray.Due to the emergence of high dimensional data,the traditional multivariate statistical methods and theories have been greatly challenged.In the high-dimensional mean value vector test,when p is greater than n,the traditional test statistics lose significance and its limit theory is no longer applies,because the traditional statistical limit theory is based on the case that the dimension P is far less than the sample size n.For this reason,many researchers have proposed corresponding statistics to solve this problem.Their research ideas are basically based on theoretical derivation,starting from the new statistics to find its mean value estimation and covariance estimation.However,in the case of high dimension,it is difficult to find out the accurate covariance estimation.In this paper,we study the problem of mean vector test based on stationary bootstrap method in high dimensional data.Through theoretical derivation,we know that the sta-tionary bootstrap is suitable for square summation of studentized column statistics.Through a simulation experiment and comparison with some previous methods,we can find that:The statistics of the sum of squares of the student column statistics based on the stationary bootstrap method has relatively good performance,because of the new statistics having relatively good experience power and experience size control,and the method avoids the complicated calculation of the covariance esti-mation of the statistics in theory.