基于典型相关分析的无约束综列协整检验  

Unrestricted panel cointegration test based oncanonical correlation analysis

作　　者：杨继生[1] 

机构地区：[1]华中科技大学经济学院,武汉430074

出　　处：《管理科学学报》2009年第4期145-152,共8页Journal of Management Sciences in China

基　　金：国家自然科学基金资助项目(70571026);教育部人文社会科学研究青年基金资助项目(08JC790043)

摘　　要：现有的综列协整检验方法均存在较强的约束条件.该文基于Johansen典型相关分析,在允许截面个体间存在协整关系和动态影响的情况下,建立了综列协整的迹检验和最大特征值检验等似然比检验统计量.为了避免似然比检验在有限样本下的分布扭曲,在无约束综列协整检验中引入了Bootstrap程序,以提高检验结论的可靠性.仿真实验结果显示,无约束综列协整检验显著优于Groen和Kleibergen的有约束检验.Curret panel cointegration test methods suffer from too strong restrictions. In this study, we propose the likelihood ratio tests for panel cointegration based on Johansen canonical correlation analysis, trace test and maximum eigenvalue test, to allow for influences by the cross sectional cointegration and dynamic influences between each other. To avoid the finite sample size distortion of likelihood ratio statistics, we introduce the bootstrap procedure to ensure the robustness of our unrestricted panel cointegration test statistic. The Monte- Carlo simulations show that our unrestricted panel cointegration test perform better than the restricted one proposed by Groen and Kleibergen .

关 键 词：综列协整 典型相关 似然比检验 综列向量误差纠正模型 Bootstrap仿真 

分 类 号：F224.0[经济管理—国民经济]