基于ESTAR模型的单位根检验--Wald统计量的研究

摘要

已有的基于ESTAR模型的单位根检验通常假设α＝0，但实践研究表明α显著。对α≠0的情况进行单位根检验，研究该情形下的 Wald检验统计量。从理论上推导了 Wald检验统计量的极限分布，并通过 Monte Carlo随机模拟，结果表明该统计量有效。%The alternative hypothesis of exponential smooth transition autoregressive (ESTAR) nonlinearity usually assumes that α=0 ,w hile the Existing tests of the unit root hypothesis is against that .However ,empirical work indicates that the estimated value of αis always greater than zero and significant .Hence ,the paper relaxes the restriction in the test regression and investigate the Wald‐type test for a unit root process against ESTAR process .The asymptotic distributions of the test statistic are derived .Results show that the Wald statistic is effective via Monte Carlo simulation .