Course description

Causality, Machine learning and Forecasting (Planned spring 2021)


Complete course description will be ready summer 2020.

We study two central problems in applied analytics: causal analysis on one side, and machine learning and forecasting techniques on the other. The aims of the two problems are complementary, and are here presented together to emphasise their differences and connections.

Course content

  • The problems surrounding analysing causality through observational data.
  • Experiments and quasi-experiments.
  • Fundamental principles of statistical learning and forecasting techniques: bias/variance trade-off, cross validation techniques and pseudo out of sample methods.
  • An overview of some central time series models and machine learning methods, and their connection to the above issues. Applied work will use Python.

Learning outcome knowledge

The student will understand the limitations of observational studies, and the types of assumptions that are required to use observational data to make causal statements. The student will understand how experiments and quasi-experiments can be used to overcome these difficulties. Knowledge of central principles in statistical learning and forecasting will be gained, and how to apply a selection of time series models and machine learning tools in practice.