由于传统求解时间序列自回归(auto-regressive,AR)模型的最小二乘方法无法顾及设计矩阵误差,现有的AR模型迭代解法难以应用协方差传播率给出较为精确的精度评定公式。将块自助采样方法引入到AR模型精度评定研究中,并在其基础上借助Siev...由于传统求解时间序列自回归(auto-regressive,AR)模型的最小二乘方法无法顾及设计矩阵误差,现有的AR模型迭代解法难以应用协方差传播率给出较为精确的精度评定公式。将块自助采样方法引入到AR模型精度评定研究中,并在其基础上借助Sieve自助法的思想,定义了顾及设计矩阵误差AR模型精度评定的Sieve块自助采样方法。根据不同的分块准则和采样策略,给出了4种方法的重采样步骤。模拟实验结果表明,精度评定的Sieve块自助采样方法能够得到比最小二乘法、经典的AR模型迭代解法更加可靠的自回归系数标准差,具有更强的适用性。同时,北斗卫星精密钟差真实案例表明,所提出的Sieve移动块自助法、Sieve非重叠块自助法、Sieve圆形块自助法以及Sieve静止块自助法的均方根(root mean square,RMS)比总体最小二乘的RMS分别减小了70.25%、78.65%、70.89%和79.24%,进一步验证了所提算法的有效性和可靠性,为时间序列AR模型的精度评定问题提供了一种采样思路。展开更多
This paper provides a stringent proof for the general test of latent variable model with time-varying risk premium developed by Ferson and Foerster(1993) and conducts a test by selecting “the size portfolio” as samp...This paper provides a stringent proof for the general test of latent variable model with time-varying risk premium developed by Ferson and Foerster(1993) and conducts a test by selecting “the size portfolio” as sample in Chinese stock market.The Block-Bootstrap method is also adopted to study the finite sample properties of GMM.The result reveals that the Block-Bootstrap simulation of GMM is robust and P-value based on asymptotic distribution tends to be underestimated.The empirical result shows that the China’s stock market can not reject the “ 1 latent variable model”.The conclusion of this paper manifests the essence of risk and return in the China’s stock market and has great significance to the policy-making.展开更多
文摘由于传统求解时间序列自回归(auto-regressive,AR)模型的最小二乘方法无法顾及设计矩阵误差,现有的AR模型迭代解法难以应用协方差传播率给出较为精确的精度评定公式。将块自助采样方法引入到AR模型精度评定研究中,并在其基础上借助Sieve自助法的思想,定义了顾及设计矩阵误差AR模型精度评定的Sieve块自助采样方法。根据不同的分块准则和采样策略,给出了4种方法的重采样步骤。模拟实验结果表明,精度评定的Sieve块自助采样方法能够得到比最小二乘法、经典的AR模型迭代解法更加可靠的自回归系数标准差,具有更强的适用性。同时,北斗卫星精密钟差真实案例表明,所提出的Sieve移动块自助法、Sieve非重叠块自助法、Sieve圆形块自助法以及Sieve静止块自助法的均方根(root mean square,RMS)比总体最小二乘的RMS分别减小了70.25%、78.65%、70.89%和79.24%,进一步验证了所提算法的有效性和可靠性,为时间序列AR模型的精度评定问题提供了一种采样思路。
文摘This paper provides a stringent proof for the general test of latent variable model with time-varying risk premium developed by Ferson and Foerster(1993) and conducts a test by selecting “the size portfolio” as sample in Chinese stock market.The Block-Bootstrap method is also adopted to study the finite sample properties of GMM.The result reveals that the Block-Bootstrap simulation of GMM is robust and P-value based on asymptotic distribution tends to be underestimated.The empirical result shows that the China’s stock market can not reject the “ 1 latent variable model”.The conclusion of this paper manifests the essence of risk and return in the China’s stock market and has great significance to the policy-making.
