两总体协方差矩阵相等性检验的功效研究.pdfVIP

摘 要 随着计算机技术的迅速发展和广泛应用,大量的数据可以被储存,这些数据 的维数往往是非常大的。在现代的许多科技领域中都有这样的大维数据,例如: 生物学中的微矩阵数据,航天飞行器数值化，金融领域的股票分析，无线网络分 析等，但是人们在处理大维随机矩阵的相关估计和假设检验问题时，许多经典的 统计方法都不适用了，并且会导致严重误差。Bai and Silverstein(2004)得到 了大维样本协方差矩阵的线性谱统计量的中心极限定理；Zheng(2012)得到了F 矩阵的线性谱统计量的中心极限定理；Zheng(2013)得到了一般F 矩阵的线性谱 统计量的中心极限定理。基于此，本文探讨了两总体协方差矩阵相等性检验的功 效。 检验的功效是衡量一个检验方法优劣的重要指标。功效是在备择假设为真 =P H 时，拒绝原假设的概率，即：检验的功效 (拒绝 0 H 1 为真)。 F 本文利用解二元非线性方程组的方法得到了 矩阵的极限谱分布的 Stieltje 变换，再利用Zheng(2013)中的理论结果求得均值和方差，进而得到了 两总体协方差矩阵相等性检验的理论power 值。 F 关键字:线性谱统计量；中心极限定理； -矩阵；功效 I Abstract The rapid development and wide application of computer techniques permits to collect and store a huge amount data, where the number of measured variables is usually large ．Such high-dimensional data occur in many modern scientific fields, such as micro-array data in biology, making digitalization of space probes, stock market analysis in finance and wireless communication networks ．However, it is found that when dealing with large dimensional data, many classical statistical procedures induce large, even intolerable errors ． Bai and Silverstein (2004) had proved the central limit theorems (CLT) for linear spectral statistics of large dimensional sample covariance matrices ．Zheng(2012) had proved CLT for linear spectral statistics (LSS) of F matrix ．Zheng(2013) had proved CLT for linear spectral statistics (LSS) of the general F -matrix ．Therefore, this paper discussed the power of testing equality of two population covarianc