This course will be delivered in French.
Forecasting in a time-series framework arises numerous issues, such as the choice of the model, the variables to be used, and the appropriate testing strategy. Recently, the collection and the availability of high-dimensional data has been posing new challenges to applied economists that classic regression approaches cannot address.
This course is designed to cover the basic topics in macroeconomic forecasting as well as advanced topics in Machine Learning for time-series analysis.
The course is split in two parts. The first part relates to linear regressions, time series models and economic forecasting. Non-linear time series models will be also introduced. The second part focuses on recent advances in machine learning and econometrics for the analysis of high-dimensional data, such as the concept of penalized likelihood, penalized regression models (Ridge, Lasso, and Elastic net), cross-validation and information criteria, and inference in penalized regressions.
Each course will be accompanied by practical sessions with Stata.
*Please note this course will be delivered in French.
This comprehensive online course is delivered through the Zoom Webinar platform and runs over a total of 9 hours, with 4 hours each day (2-hours in the morning and 2-hours in the afternoon). Additional time is available at the end of the final session for open Q&A's.
Presentation of the models, estimation, tests and forecasting. Models to be considered: Linear regression, Auto-Regressive Distributed Lags and Non-linear time series models.
Empirical applications with data using Stata.
Introduction to penalized likelihood and Machine Learning in time-series econometrics. Review of fundamental penalized regressions (Ridge, Lasso, Elastic-Net). Selection of penalization parameters.
Empirical applications (estimation, forecasting) with cross-section and time-series data using Stata..
The number of delegates is restricted. Please register early to guarantee your place.