摘要
根据备择假设成立时均值是否为零,讨论了三类单位根检验统计量的检验功效。理论研究表明:在大样本下,检验功效只与检验类型有关,与均值取值无关。蒙特卡洛模拟表明:在有限样本下,均值会影响检验功效,当均值为零时,第一类DF检验功效最高;当均值大于零时,在绝大多数场合下,第二类DF检验功效最优。因此,递归均值调整单位根检验功效仅在有限条件下最大。
This paper discusses the testing power for three kinds of unit root statistics according to whether the mean of series is zero or not when the alternative hypothesis is true .The theoretical research shows that the testing power only depends on the type of statistics and has nothing to do with the mean size under the condition of large samples ,which is contrary to the conclusion from the Monte Carlo Simulation that the value of mean do affect the power when the sample is finite .The testing power of first type DF is the highest when the mean is zero .When the mean is greater than zero ,the power of second type DF is much higher than that of recursive mean adjustment in most cases .Therefore ,the effectiveness of recursive mean adjustment is subjected to limited conditions .
出处
《统计与信息论坛》
CSSCI
2014年第11期3-10,共8页
Journal of Statistics and Information
基金
国家社会科学基金项目<基于Bootstrap方法下单位根检验研究>(13BJY011)
教育部人文社会科学基金项目<基于灰技术的区域能源-经济-环境复杂系统研究>(11YJC630034)
关键词
递归均值调整
单位根
功效
蒙特卡洛模拟
recursive mean adjustment
unit root
power
Monte Carlo simulation
