协整模型中门限协整的一种辨识方法

Identification Method of Threshold Cointegration in Cointegrating Model

考虑非线性、非平稳时间序列中门限协整模型的辨识问题,给出协整模型属于线性协整还是门限协整的一种判断方法。给出门限协整回归模型的定义,构造了辨识线性协整和门限协整的SupW统计量,并得到了SupW统计量的极限分布。给出逼近极限分布的Bootstrap方法,计算渐近p-值确定协整模型是线性协整还是门限协整。模拟计算和实例计算证明了该辨识方法的有效性,具有重要的应用价值。

For modeling nonlinear and nonstationary time series, an identification method was proposed to detect threshold cointegration in cointegrating regression model. A SupW statistic was given and its null asymptotic distribution was derived. The residual-based block bootstrap was presented for approximating to limiting distribution and computing asymptotic p-value. The effectiveness of this method was investigated by the finite sample simulations. The results show that the empirical size of the method is close to the normal one and the method succeeds in distinguishing threshold cointegration from linear cointegration.