基于序列与逆序列最小Wald统计量的通用STAR模型平稳性检验法  

A General Stationary Test Method of STAR Model Based on the Minimum of Wald Statistics of a Series and Its Adverse Series

作　　者：刘田[1] 谈进[2] 

机构地区：[1] 西南财经大学统计学院 [2] 西南财经大学经济信息工程学院

出　　处：《统计研究》2014年第3期99-105,共7页Statistical Research

摘　　要：本文提出一种通用非线性单位根检验方法,使用待检序列及其逆序列的Wald统计量的最小值作为检验统计量,将Kapetanios等人提出的受限条件下ESTAR模型非线性单位根检验推广到非0位置参数的情形,也可应用于LSTAR或其他可能的STAR模型、TAR模型或传统的线性AR模型的平稳性检验。并且,推导了检验统计量的极限分布,仿真了检验水平与功效。结果表明,本文提出的检验方法对数据生成过程有广泛的适应性,并且在大多数情况下都能获得较其他方法更佳的检验功效。This paper proposes a general nonlinear unit root test method, using the minimum of Wald statistics of a series and its adverse series as test statistic. The method generalizes the nonlinear stationary test method against constrained ESTAR model proposed by Kapetanios et al. to nonzero position situation, at the same time, it can be applied to test first-, second-order LSTAR model, or other possible smooth transition autoregressive model, or threshold autoregressive model and traditional linear AR model. We derive the limiting nonstandard distribution of the proposed test and simulate the test powers. The results suggest that the new test has wide adaptability to data generating process and is generally superior in terms of power.

关 键 词：平滑转移自回归模型 非线性平稳性检验 伪检验 蒙特卡洛仿真 

分 类 号：O12[理学—数学]