中文摘要：

本文将协整检验由传统的线性协整检验扩展到线性协整检验和闽值协整检验，并在Enders和Granger（1998）方法的基础上提出了一个新的检验协整是否存在的Sup—F和Sup—F。统计量。通过MC仿真研究发现：在线性协整下，ADF方法比Sup-F法具有更高的检验势，但在“持久性”较强时，Sup—F检验比ADF检验法具有更高的检验势；Sup—F统计量在Three-Regime的闽值协整检验中比ADF法有更高的检验势；Sup—F’在检验协整（包括线性协整和闽值协整）时都具有较低的检验势；随着在不同Regime中自回归系数差距的增大（非对称程度增大），Sup-F统计量的检验势提高很快，且比ADF法的检验势高。

英文摘要：

In this paper, we think that the co-integration test involves the tests of linear co-integration and threshold co-integration. Accordingly, we proposes Sup-F and Sup-F test for the presence of linear co-integration or threshold co-inte- gration based on Enders-Granger method （1998）. A set of Monte-Carlo simulation shows that： First, ADF test power is greater than that of Sup-F method under linear co-integration. However when the regression error has strong persistence, Sup-F test power exceeds the power of ADF method. Second, Sup-F test has greater power than that of ADF method under the Three-Regime threshold co- integration. Third, whether the co-integration is linear or not, the Sup-F test has low test power. Fourth, in the presence of a substantial amount of asymmetry, the Sup-F test power rises faster than that of the ADF method.