单位根过程联合检验的Bootstrap研究  

Bootstrapping Research on Joint Test for Unit Root Processes

作　　者：陶长琪[1,2] 江海峰[3,1] 

机构地区：[1]江西财经大学信息管理学院 [2]中国数量经济学会 [3]安徽工业大学

出　　处：《统计研究》2013年第4期106-112,共7页Statistical Research

基　　金：国家自然科学基金(71073073;71273122);国家社会科学基金(12BJY014);教育部人文社会科学研究一般项目(11YJA630169)的资助

摘　　要：本文以Ferretti和Romo的Bootstrap方法为基础进行拓展完成三种联合检验,并从理论上证明了Bootstrap方法的有效性;使用蒙特卡洛模拟技术比较了Bootstrap检验与临界值检验的效果。模拟表明,在误判率上,Bootstrap方法下三个检验量的误判率分别为2.22%、3.70%和0.00%,而临界值的误判率分别高达22.22%、11.11%和15.56%;在精确程度上,Bootstrap方法的精度分别是临界值方法的11.25倍、26倍和6.5倍。模拟表明了本文构造的Bootstrap检验方法可以替代临界值方法,特别在小样本下,Bootstrap方法的优势更为明显。Based on the Bootstrap method proposed by Ferretti and Romo, the article extends it to perform the three joint test and the strict theoretical proof for its effectiveness is also given. Then, Monte Carlo simulation is used to compare the effect of the method between Bootstrap and the critical values. The results from simulation show that the rate of erroneous judgment for three joint tests are 2. 22% , 3.70% , 0. 00% for Bootstrap method and high up to 22. 22% , 11.11% , 15.56% for the critical value method separately, as for the exactness, the Bootstrap method is 11.25, 26 and 6.5 times higher than that of the critical value method for the three joint tests separately. These results testify that the Bootstrap method constructed in this paper can not only replace the critical value technology but also has the obvious advantage over it in small sample size.

关 键 词：单位根过程 联合检验 Bootstrap检验 Bootstrap不变原理 

分 类 号：O212[理学—概率论与数理统计]