TIME SERIES ECONOMETRICS

Professor: Stanislav Anatolyev

The course is devoted to the modern time series analysis. We will study various model selection procedures, and get acquainted with general principles of constructing and evaluating forecasting models. We will review popular linear models of the conditional mean dynamics such as ARs and VARs, and then concentrate on nonlinear models, especially those exhibiting regime switching. We will also review such issues as stationarity, trends, unit roots and cointegration. After that, we will turn to models of conditional variance as represented by the ARCH class. Finally, we will study some special methods like the analysis of structural breaks, retrospection and monitoring, and modeling of ultra-high frequency financial data.

COURSE OUTLINE

I. Modeling methodology and model selection

+        Modeling the mean and modeling the variance.

+        Model selection: diagnostic testing, information criteria and prediction criteria.

+        General-to-specific and specific-to-general methodologies. Data mining.

+        Forecasting principles. Forecast evaluation and testing for predictability.

II. Modeling the mean

+        Linear AR models: properties, estimation, inference, forecasting.

+        Nonlinear time series modeling of the mean: SETAR, STAR, MSW models.

+        Linear VAR models: properties, estimation, analysis and forecasting. Nonlinear VAR.

+        Stochastic and deterministic trends, unit roots. Spurious regression and cointegration.

III. Modeling the variance

+        The class of ARCH models: properties, estimation, inference and forecasting.

+        Extensions: IGARCH, ARCH-t, ARCD, ARCH-in-mean, multivariate GARCH.

IV. Special topics

+        Identifying and testing for structural breaks.

+        Retrospection and monitoring.

+        High frequency data models: ACD, UHF-GARCH.