摘要
文章研究平稳的平滑转移自回归过程之间的虚假回归问题。通过推导最小二乘回归估计量及其对应的t统计量的极限分布,发现:标准的t检验流程中的t统计量并不趋于标准正态分布,其极限分布依赖于模型参数,从而导致了虚假回归的可能。采用蒙特卡洛模拟研究了有限样本下数据生成过程的各项参数对虚假回归的影响,研究表明:虚假回归现象也可能普遍存在于平稳变量之间,为此,在做统计推断时,考虑平稳变量的具体特征是必要的。
This paper examines the problem of spurious regression between the stationary smooth transition autoregressive processes.By deriving the limiting distributions of the estimates and t-statistics,we find that the t statistics of the standard t test cannot converge to the standard normal distribution,and its distribution depends on the parameters of the data generating processes,which leads to the possibility of spurious regression.We analyze the impact of the parameters of data generating processes and the sample size on spurious regression via Monte Carlo simulations.The evidences of Monte Carlo simulations support our theory.This paper implies that the phenomenon between stationary variables may a common existence.Therefore,it is necessary to consider the specific characteristics of stationary variables in statistical inference to avoid the spurious regression phenomenon.
出处
《统计与决策》
CSSCI
北大核心
2017年第6期15-19,共5页
Statistics & Decision
基金
中央高校基本科研业务费专项资金资助项目(15LZUJBWZY118
15LZUJBWZY097)
关键词
平滑转移
虚假回归
蒙特卡洛模拟
smooth transition
spurious regression
monte carlo simulation