基于分形理论的LSTAR模型检验

The Test in LSTAR Model Based on Fractal Theory

针对非线性时间序列的单位根检验和非线性检验中存在的问题,引入分形理论中的Hurst指数进行研究,以LSTAR模型为代表,讨论了相关样本性质。通过蒙特卡洛模拟实验证实,只有当非平稳特征十分明显时,DF检验才能检测出单位根。Hurst指数改善了单位根检验功效,且作为一种非参数检验方法,具有较好的稳健性。

LSTAR model as the representative to study the properties of relevant sample by the introduction of the Hurst exponent in the fractal theory, for the problems of the unit root test and nonlinear test in the nonlinear time series. The Monte Carlo simulation tests confirm that the effect of the propose unit root test statistic is significantly higher than that of DF unit root test, and only when the non-staionary feature is very obvious, DF tests can detect the unit root. Hurst exponent improved power efficiency of unit root test, and has good stability as a method of Nonparametric Tests.