Time Series Analysis in Data-Rich Environments

I. Syllabus:

The aim of the course is to make you acquainted with techniques allowing you to deal with the situation where the number of available predictors in a time series forecasting problem is comparable with the number of observations, such that the usual OLS estimator of the predictive regression is highly unstable and the resulting forecasts highly imprecise. The rst part of the course presents usual evaluation and comparison techniques for time series forecasts. The second focuses on the case with a large number of predictors, and the third deals with the case where a (not so large) number of predictors is observed at a higher frequency than the series to be forecast (e.g. daily vs. monthly or quarterly). While the rst part is more lecture-style, the latter two are based on discussing relevant articles, as will be the exam. When it comes to technical aspects, we do not go too much into details.

  • Forecast evaluation and comparison
  • Factor-based forecasting in the many predictors case
  • Predictor selection
  • MIDAS (mixed data sampling) regressions
  • Mixed-frequency VARs (if time permits)


II. Prerequisites:


III. Exam:

  • oral exam
  • Examina Day Date Time Room


    IV. Downloads:


    V. Literature:

    • Background notes as pdf file.
    • Elliott, G. and A. Timmermann (2008). Economic Forecasting. Journal of Economic Literature 46, 3-56.
    • Relevant research articles tba.

      VI. Lecture:

      Day Time Room Lecturer Date UnivIS


      VII. Voluntary tutorial (paper/pen/PC-LAB):

      Instructor Room Day Time Dates


      VIII. Registration:

      Access to the computer lab requires a one-time registration with a Stu-Account.