时间序列分析（36学时）课程简介

时间序列分析是统计学的一个非常重要的分支，是经典计量经济学分析方法的延伸，是以概率论与数理统计为基础、计算机应用为技术支撑，迅速发展起来的一种应用性很强的科学方法，用以分析、探索社会、经济、金融数据的动态结构和发展变动规律，进而对其未来状态进行预测控制。

时间序列分析（36学时）主要讲授经典的时间序列模型，包括平稳时间序列模型、非平稳时间序列模型、向量/多变量时间序列模型、波动性模型等部分的理论及其实际应用，另外还包括预测理论。平稳时间序列模型部分讲授AR(p)模型、MA(q)模型与ARMA(p, q)模型的基本概念、自相关函数、部分自相关函数、自相关性检验方法（包括自相关图Q统计量、D-W统计量、LM统计量），以及平稳时间序列模型的估计、识别与实际应用；非平稳时间序列模型部分讲授非平稳时间序列的单位根检验法、确定性趋势模型、随机性趋势模型、去除趋势的方法与实际应用；向量/多变量时间序列模型部分讲授协整理论、向量自回归（VAR）模型、结构向量自回归（SVAR）模型，以及向量误差修正模型（VECM）的估计、统计推断与实际应用；波动性模型部分讲授ARCH模型、GARCH模型、TGARCH模型、EGARCH模型等GARCH类模型的属性、估计、检验与实际应用。预测理论部分讲授静态预测、动态预测、以及预测理论的基本概念和预测准确性的度量指标。

使用教材：张成思. 金融计量学：时间序列分析视角. 北京：中国人民大学出版社，2012.

Time series analysis (36 class hours)

Time series analysis is a rapidly developing application-oriented scientific method which is an important branch of statistics and an extension of classic econometric analysis method, based on probability theory and mathematical statistics, and the application of computer technology to support. We can make use of the method to analyze and explore the dynamic structures and the law of development in the social, economic and financial data, and then predict and control the future state.

Time Series Analysis (36 class hours) mainly interprets classic time series models which includes Stationary Time Series Models, Non-Stationary Time Series Models, Vector Time Series Models, Volatility Time Series Models and Prediction Theory with their applications.

Stationary Time Series Models branch focuses on explaining AR(p) models, MA(q) models, ARMA(p,q) models with their basic properties, autocorrelation and partial correlation function, serial correlation tests (for example correlogram Q statistics, Durbin-Watson statistics, and Breusch-Godfrey serial correlation LM tests), model estimation and identification with their applications.

Non-Stationary Time Series Models branch will mainly introduce unit root tests for non-stationary time series, general non-stationary time series models (both deterministic trend models and random trend models), and detrending methods with their applications.

Vector/Multivariate Time Series Models branch focuses on cointegration theory, model estimation and statistical inference of vector autoregression models (VAR), structure vector autoregression models (SVAR) and vector error correction models (VECM) with their applications.

Volatility Models branch analyzes the various volatility models based on volatility clustering in micro-financial data. ARCH models, GARCH models, and asymmetric GARCH models (both TGARCH models and EGARCH models) with their properties, model estimation and testing for ARCH effect will be particularly introduced.

The prediction method includes static prediction and dynamic prediction. Prediction Theory branch will introduce the basic definition of prediction theory, and the measurement indices of prediction accuracy such as MSE, RMSE, MAPE and so on.

《时间序列分析》课程教学进度计划