非线性时间序列模型研究及实证分析

发布时间：2018-03-04 21:36

本文选题：非线性时间序列分析　切入点：叠合模型　出处：《电子科技大学》2014年硕士论文　论文类型：学位论文

【摘要】：金融市场中人们往往关注金融产品的收益率,而收益率序列往往表现出波动性集群现象,因而具有变方差等非线性特征。而非线性时间序列模型,包括参数模型如ARCH模型、TAR模型,非参数模型如AAR、低阶NAR模型等,在刻画时间序列的非对称性、长记忆性、变方差性、非线性自相依性等方面,具有明显的优越性。因此应用非线性时间序列分析方法研究金融数据,能够给金融计量学学者和金融投资者提供更为广阔的视角。论文先简要介绍了常见的非线性时序模型：讨论了非线性时序分析中非参数方法“维度祸患”产生的原因；总结介绍了基于图示方法与似然比检验的时间序列非线性特征识别。本文讨论了基于广义似然比检验的时间序列非线性自相依性的识别方法；解释了将不同非线性模型进行叠合的原因,并且讨论了一些叠合模型(包括NAR-ARMA模型、NAR-AR-GARCH模型等)的参数估计。论文应用非线性时序分析方法研究美国大豆期货收益率。利用基于图示的方法与基于广义似然比检验的方法研究了美国大豆期货收益率序列的内部动态规律,以及收益率与其外部影响因素之间的关系。结果表明：美豆期货收益率序列存在显著的非线性自相依结构；美国大豆生长优良率以非线性的方式显著影响其期货收益率;存在其他因素以非线性方式显著的影响美豆期货收益率。最后对美豆期货收益率分别建立了参数叠合模型与半参数叠合模型：AR-GARCH模型与NAR-AR-ARCH模型,进一步计算拟合优度,结果表明NAR-AR-GARCH模型远比AR-GARCH模型适合数据。这与收益率非线性自相依特征相一致,同时也在一定程度上说明了半参数非线性叠合模型在金融时间序列波动性分析应用中的有效性。
[Abstract]:In the financial market, people tend to pay attention to the return rate of financial products, and the return sequence often shows the phenomenon of volatility cluster, so it has nonlinear characteristics such as variable variance, and nonlinear time series model. It includes parameter model such as ARCH model, nonparametric model such as ARCH model, low order NAR model and so on. It describes the asymmetry of time series, long memory property, variableness of variance, nonlinear self-dependence and so on. Therefore, using nonlinear time series analysis method to study financial data, It can provide a broader perspective for financial metrologists and financial investors. Firstly, this paper briefly introduces the common nonlinear time series models: discusses the causes of the "dimensionality disaster" in the non-parametric method of nonlinear time series analysis; The nonlinear feature recognition of time series based on graphical method and likelihood ratio test is summarized. In this paper, the identification method of nonlinear self-dependence of time series based on generalized likelihood ratio test is discussed. The reasons why different nonlinear models are superimposed are explained. The parameter estimation of some superposition models (including NAR-ARMA model, NAR-AR-GARCH model, etc.) is discussed. The nonlinear time series analysis method is applied to study the yield of American soybean futures. The graphical method and the generalized likelihood ratio are used to study the yield of soybean futures in the United States. The method of testing is used to study the internal dynamic law of the yield series of American soybean futures. The results show that there is a significant nonlinear self-dependent structure in the yield sequence of American soybean futures, and the growth rate of American soybean significantly affects its futures yield in a nonlinear manner. There are other factors that influence the yield of soybean futures significantly in a nonlinear way. Finally, the parametric superposition model and semi-parametric superposition model and NAR-AR-ARCH model are established to calculate the goodness of fit. The results show that the NAR-AR-GARCH model is far more suitable for the data than the AR-GARCH model, which is consistent with the characteristics of the nonlinear self-dependence of the rate of return. At the same time, the validity of the semi-parametric nonlinear superposition model in the analysis of the volatility of financial time series is explained to a certain extent.
【学位授予单位】：电子科技大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：F830.9;F224

【相似文献】