预测回归模型序列相关性的Jackknife经验似然比检验  

A Jackknife Empirical Likelihood Ratio Test for Serial Correlation of Predictive Regression Model

作　　者：范雅文 刘小惠[1,2] FAN Yawen;LIU Xiaohui(School of Statistics,Jiangxi University of Finance and Economics,Nanchang 330013;Research Center of Applied Statistics,Jiangxi University of Finance and Economics,Nanchang 330013)

机构地区：[1]江西财经大学统计学院,南昌330013 [2]江西财经大学应用统计学研究中心,南昌330013

出　　处：《系统科学与数学》2019年第9期1471-1485,共15页Journal of Systems Science and Mathematical Sciences

基　　金：国家自然科学基金(1601197,11971208);中国博士后科学基金(2016M60051120,17T100475);江西省自然科学基金会(20171ACB21030,2018ACB21002);江西财经大学学生科研课题(xskt19393)资助课题

摘　　要：预测回归模型是计量经济学中的重要模型,而相关的序列检验问题在文献中尚未被提及.考虑到序列相关性检验在模型实证中的重要性,文章基于Jackknife经验似然法,构造了该模型的序列相关检验统计量,并在一定条件下推导了零假设成立时检验统计量的渐近分布,分析了它在局部备择假设下的功效情况,最后通过随机模拟和实际数据例子验证了该检验方法的有限样本性质.Predictive regression model is a very important model in econometrics.However,the issue of its serial correlation test has not been mentioned in the literature.Considering the importance of the serial correlation test in the empirical study,this paper considers a jackknife empirical likelihood based serial correlation test.The limit distribution,as well as the power property at local alterative hypotheses,of this statistic has been derived under certain conditions.A few examples are also given to illustrate its finite sample performance in terms of size and power based on some simulated and real data sets.

关 键 词：预测回归模型 序列相关检验 JACKKNIFE 经验似然 

分 类 号：F22[经济管理—国民经济]