Brief summary of the course
The course consists of four main parts. In the first part, we focus on the objectives of forecasting. We discuss different types of forecasts and approaches to measuring their goodness. Afterwards, we focus on the main differences between modelling cross-sectional and time-series and on the characterization of time series (ACF, PACF, etc.). The second part is devoted to basic approaches to time series decomposition and forecasting. Here we first discuss smoothing approaches to trend extraction (MA, LOESS) and then forecast oriented tools such as EWMA and Holt-Winters methods of decomposing and forecasting TS with season and trend. In the third part we begin with the classical ARMA modelling, but extend this setting to integrated and seasonal data. We discuss estimation, testing and forecasting in full detail. The fourth part of the course gives the students the opportunity to learn several specific areas of time series modelling. We begin with non-linear GARCH-type processes that are of particular importance in modelling financial markets (Nobel Prize in Economics). Then we proceed to two data-driven problems such as outlier detection and detection of structural breaks. Here we distinguish between ex-post and online monitoring. Finally, we move to multivariate time series models and focus on VAR processes. To reflect the dynamic environment of real data we discuss state-space models and regime-switching models. In the last part, we link this course to the Econometrics course and consider models for panel data.
Course topics
- Basics of forecasting and time series analysis
- Forecasting using regression
- Forecasting cross-sectional data
- Regression for time series data
- Forecasting using spline/trigonometric regression
- Time series decomposition
- Exponential smoothing (EWMA, Holt, Holt-Winters, Croston)
- SARIMA modeling
- Special topics in time series and forecasting
- ARCH/GARCH: models with conditional volatility
- Outliers in time series
- Structural breaks in time series
- Multivariate time series
- State-space models and Kalman filtering
- Panel data
Prerequisites
- Statistics and Econometrics
- Basic knowledge in analysis, statistics and econometrics is of advantage.