时间序列非线性存在性检验方法及其应用  

Nonlinear Existence Test Method in Time Series and Its Application

作　　者：舒晓惠[1] 宋金奇[2] Shu Xiaohui;Song Jinqi(School of Economics,Huihua University,Huihua Hunan 418000,China;College of Science and Technology,Jiangxi Normal University,Nanchang 330027,China)

机构地区：[1]怀化学院经济学院,湖南怀化418000 [2]江西师范大学科技学院,南昌330027

出　　处：《统计与决策》2018年第16期70-73,共4页Statistics & Decision

基　　金：国家社会科学基金青年项目(11CTJ003)

摘　　要：文章探讨了时间序列中被忽略的非线性,并给出了非线性存在性检验的组合检验方法：Q检验、BDS检验、BP加强型神经网络、加强型高斯小波神经网络以及加强型墨西哥帽小波神经网络方法。进一步,对于金融理论中货币需求函数的稳定性检验问题,应用组合检验方法实证研究发现,对于构建的货币需求函数的各变量序列,均存在不同程度的被忽略的非线性,因此有必要引入非线性研究方法。这也为经济变量序列存在的非线性特征提供了一个重要证据,要求我们在研究经济内在规律中考虑非线性协整理论与方法的应用。This paper discusses the neglected nonlinearities in time series and gives the combined testing methods for nonlinear existence test,namely Q test, BDS test, BP-enhanced neural network, enhanced Gaussian wavelet neural network and enhanced Mexican hat wavelet neural network method. And then for the stability test of monetary demand function in financial theory, the paper makes an empirical study by use of combined testing method to find that there exist different degrees of neglected nonlinearity in the variable sequence of the constructed monetary demand function, hence the need to for the non-linear research method to be introduced into the stability of demand function. This also provides an important evidence for the nonlinear characteristics existing in economic variables sequence, requiring us to consider the application of nonlinear co-integration theory and method in the study of internal laws of economy.

关 键 词：非线性 Q检验 BDS检验 神经网络 

分 类 号：O212[理学—概率论与数理统计]