针对线性以及非线性协整检验存在模型参数过多、小样本条件下检验功效偏低的问题,提出基于非参数ACE变换的贝叶斯非线性协整VAR模型,运用ACE算法进行变量变换,垒;合参数的完全条件分布设计Gibbs抽样方案,进行贝叶斯非线性协整检验,并利用MonteCarlo仿真研究了贝叶斯非线性协整方法的检验势,发现贝叶斯非线性协整比经典Johansen法具有更高更稳健的检验势;同时,对中国城市和农村居民消费价格指数序列进行实证分析.研究结果表明:贝叶斯非线性协整方法解决了模型中参数过多、小样本条件下检验功效偏低的问题,提高了估计的精确度和检验的准确性.
Two methods of identifying cointegrating relationship between variables are commonly used: linear cointegration analysis and nonlinear cointegration procedure. Because the test power of all these methods is poor in small sample, this paper proposes a Bayesian method for conducting inference about nonlinear cointegrating vector autoregressive model based on ACE transformation. The idea of ACE algorithm is adopted for the nonlinear transformation of the variables, with Gibbs sampling being used to carry out Bayesian nonlinear cointegration identification. Numerical results are produced via a combination of Monte Carlo simulation, from which we find that Bayesian nonlinear cointegration test is noticeably more powerful and robust than Johnnsen' s cointegration analysis. Finally, through the empirical application in Chinese urban and rural consu nption price indices, the usefulness of this Bayesian method is demonstrated. As a result, it shows that Bayesian nonlinear cointegration solves poor-power problem in small sample and improves the precision of cointegration test.