关键词: LSTAR-GARCH模型, 单位根检验, 极大似然估计, 蒙特卡洛模拟

Abstract: The unit root test of LSTAR model often ignores its time-varying conditional variance, in fact, for many economic variables, especially the financial variables, after LSTAR model set up, we often found the conditional variances exist GARCH effects. In view of the problem of stationarity test of LSTAR-GARCH model, this paper constructs the test statistics tNG, then on the basis of the maximum likelihood estimation, derived the asymptotic distribution of tNG, asymptotic critical value of the statistic is obtained by the Monte Carlo simulation method, and on the basis, studies the test power. Comparing with the tNG test proposed by Liu and Zhang(2009)、the tLG test proposed by Ling et al.(2003) and the standard Dickey-Fuller test, we found that our proposed test has the best test power.

Key words: LSTAR-GARCH Model, Unit Root Test, Maximum Likelihood Estimation, Monte Carlo Simulation