Forecasting and Prediction Methods
Are you looking to improve your forecasting and prediction skills? Would you like to improve or brush-up your knowledge of time-series econometrics and data science? This certificate covers all practical aspects of forecasting and prediction of timeseries data and follows a hands-on approach. We focus on the practical design and estimation of dynamic econometric models, forecasting techniques, dynamic policy analysis, modeling stationary and non-stationary time-series, working with panel data, and state-space models. The modules can be completed using MATLAB, EViews, python or R. The certificate of Time-Series Econometrics and Data Science provides participants with an understanding of how to implement and use advanced time-series techniques to produce high-quality forecasts and predictions.
We cover a wide range of applications in business, economics, finance and social sciences. This certificate will allow you to gain experience in analyzing client data for forecasting product demand and company sales. We also show how these predictions can be used for setting optimal predictive pricing policy, selecting best marketing strategies, optimizing inventories, and developing data-driven business strategies. We will also discuss the forecast an prediction of aggregate variables such as global demand, and country-wide inflation. We show how these models can be used for policy analysis by governments and central banks.
Module 1 (one full day of training): Time-Series Forecasting and Dynamic Econometrics. It covers the basics of time-series analysis and introduces participants to dynamic models for forecasting and prediction. Participants will learn how to design, implement and estimate dynamic models for stationary data. Additionally, we will show how to ultimately use these models to produce quality forecasts and predictions. We will cover a wide range of applications in business, economics, finance and social sciences. This module will allow you to gain experience in analyzing client data for forecasting product demand and company sales. We also show how these predictions can be used for setting optimal predictive pricing policy, selecting best marketing strategies, optimizing inventories, and developing data-driven business strategies. We will also discuss the forecast and prediction of aggregate variables such as global demand, and country-wide inflation. We show how these models can be used for policy analysis by governments and central banks.
Module 2 (one full day of training): Forecasting Methods for Non-Stationary Time-Series. This module focuses on the design, implementation and estimation of dynamic models for non-stationary data. Most economic and financial data is non-stationary in nature. Without a careful understanding of the non-stationary nature of time-series, data scientists obtain invalid results that translate into inaccurate and erroneous forecasts and predictions. This module introduces participants to the fundamental problem of spurious regression. By the end of this module participants will understand why the forecasting and prediction of non-stationary time-series data is so challenging and requires the use of special techniques and methods. Participants will learn how to avoid invalid and inaccurate results when dealing with non-stationary data. Ultimately, participants will know how to properly model, predict and accurately forecast non-stationary time-series. We will cover a wide range of applications in business, economics, finance and social sciences.
Module 3 (one full day of training): Econometric Methods for Panel Data. This module focuses on prediction and forecasting by using large panel data sets. Panel data sets can be considered as a collection of time-series data on multiple cross-section units. Cross-section units can be countries, companies, clients, stocks, geographical regions etc. Prediction and forecasting using panel data sets is becoming more and more popular as the data availability increases in many fields. Using panel data sets provides a greater ability to explain heterogeneity across individuals over time. It is more informative and provides more accurate inference of model parameters since it utilizes bigger data sets. Businesses, financial institutions, central banks are frequently encountering panel data estimation exercises in their operations. In this module, we will cover methods of prediction and forecasting that are developed for panel data models. We will discuss methods that allows for dependence across units, a phenomenon called “cross-sectional dependence” in panel data models. This will allow us to use more realistic models and obtain more accurate predictions and forecasts. Furthermore, we will discuss various examples of practical implementations of the methods. Finally, we will illustrate the methods with a panel data set application that involves financial and economic variables such as interest rates, house prices, inflation, stock returns. By the end of this module, participants will become familiar with methods to analyze large panel data models with cross-sectional dependence. They will learn how to implement these methods while working with panel data sets in empirical applications.
Module 4 (one full day of training): Forecasting Time-Series with State Space Models. This module treats the econometric challenges of forecasting financial time series using the latest Kalman filter methods. We provide introductions to general forecasting methods, time series models in state space form, the Kalman filter and related methods for signal extraction and forecasting. The methods are illustrated for a range of financial time series including asset returns, risk measures, interest rates and the yield curve, house prices, inflation, etc. The introductory treatment makes it a very accessible module with only a basic knowledge of statistics and econometrics required. At the same time, the methods are discussed in detail. Hence this module presents to you the opportunity to enhance your econometrics skills in forecasting financial time series. The module will consist of a mix of learning new technical methods, illustrations in finance, hands-on computer practice of implementation, and empirical analysis and forecasting of financial time series data. Specific attention will be given to the embedding of forecasting in data science, predictive analytics and machine learning.
The certificate is composed of four one-day modules:
- Time-Series Forecasting and Dynamic Econometrics
- Forecasting Methods for Non-Stationary Time-Series
- Econometric Methods for Panel-Data
- Forecasting Time-Series with State Space Models
Each module is a one-day course (9am to 5 pm) and can be taken separately. Successfully completing all modules will grant you the certificate diploma in "Time-Series Econometrics and Data Science".