摘要
本文从初始值的视角研究其对随机系数面板数据单位根联合LM检验稳定性的影响。推导当初始值不是依概率有界的随机变量而是渐近不可忽略的变量时联合LM检验统计量的渐近分布,并发现联合LM检验不再服从其原分布,且其与初始值有关,表明联合LM检验统计量的渐近性质不稳定。蒙特卡洛模拟结果显示,在有限样本情形下,初始值可能会导致联合LM检验统计量出现较严重的水平扭曲现象,即联合LM检验的统计性质不稳定,易受初始值影响。
This paper considers whether the effects of initial value will affect power of the random coefficient panel data unit root joint LM test. The first part of this work is theo- retically derived the asymptotic distribution of the joint LM test, when the initial value is as- ymptotically cannot be ignored, the theoretical study shows that the asymptotic distribution of the test is related to the initial LM value, and no longer subjects to its original distribu- tion. The second part of this paper studies whether the initial value makes a difference to the finite sample effects of the joint LM test by Monte Carlo simulation, and the simulation re- suits show that the initial value
出处
《数量经济技术经济研究》
CSSCI
北大核心
2016年第10期154-160,F0003,共8页
Journal of Quantitative & Technological Economics
基金
中央高校基本科研业务费项目"时间序列前沿研究团队"(JBK150501)
中央高校基本科研业务费项目"金融大数据研究基地"(JBK140403)的资助
关键词
初始值
稳定性
渐近分析
随机系数
水平扭曲
Initial Value
Size Distortion 1s easy to cause serious size Stability
Asymptomatic distortions of the joint LM test. Analysis
Random Coefficient
