基于互相关系数的二元随机过程伪回归探讨  

Discussion of Spurious Regression in Binary Stochastic Processes Based on Correlation Coefficient

作　　者：刘薇[1] 常振海[1] 张德生[2] 

机构地区：[1]天水师范学院数学与统计学院,甘肃天水741001 [2]西安理工大学理学院,西安710054

出　　处：《山西大学学报（自然科学版）》2015年第3期459-464,共6页Journal of Shanxi University(Natural Science Edition)

基　　金：国家自然科学基金(51379172);天水师范学院中青年教师科研资助项目(TSY201204)

摘　　要：通过对样本互相关系数分布的讨论,特别是其方差的变化,分析了伪回归产生的原因.结论是二元随机过程X,Y存在自相关性或非平稳等情况,导致样本互相关系数的方差变大,从而以一定的规则来判定时产生错判;其次,针对伪回归产生的原因,提出了避免伪回归现象的办法;通过模拟分析和实证分析说明了方法的可行性及结论的正确性.The causes of spurious regression are analyzed through the discussion of distribution of the sample correlation coefficient,in particular its variance changes.The results show that the existence of autocorrelation or non-stationary in Binary Stochastic Processes X,Yleads the variance of the sample correlation coefficient to become large.So the erroneous judgments are generated by certain rules.Moreover,the method to avoid the spurious regression is proposed for the causes of spurious regression;The feasibility of method to avoid spurious regression and correctness of conclusions are illustrated by simulation analysis and empirical analysis.

关 键 词：二元随机过程 伪回归 样本互相关系数 自相关 非平稳 

分 类 号：O212[理学—概率论与数理统计]