Data Analysis: Prediction and Causality

The course is part of Economics specialization (Schwerpunkt Economics) or other options following your relevant examination regulations. The scope of the course is 9 ETCS (9 LP). Further details may be found in the module manual.

The course is held twice a week, with one session lasting two teaching hours (1.5 h) and the other three teaching hours (2.25 h). The three-hour session alternates between a lecture and a tutorial.

The course requires basic stochastic (random variables, distributions, moments), inference (estimators, sampling distribution, bias, test-statistics, asymptotics), regression (linear model, OLS) and causality (experiments, selection effect, omitted variable bias) knowledge. The tutorials are based on R / RStudio. Basics skills in R are therefore required. Students that do not bring those prerequisites must invest an additional effort for this course.

Further this course contains submissions of problem sets that are a mandatory requirement for writing the exam. This problem sets are solved in groups up to 5 students using R / RStudio. Additionally groups may participate in projects to gain a 0.3 bonus for the exam.

Students in this course learn to critique empirical results in academia and business as well as doing their own analysis.

We focus on a very limited set of models (regression, mostly linear regression). These models are often sufficient and always a good starting point, but in addition it allows us to focus on

• the details of the specifications (e.g. what variables to include on the right hand side?)
• the right statistics to look at (e.g. P-value, R2, CV-Error?)
• the correct interpretation of the statistics (causal or not?)
• the limits of any analysis.

The lecture teaches different models and methods, that students apply afterwards in the mandatory problem sets. The solution of the problem sets is discussed in the tutorial. In addition, the students will prepare and present a number projects in the lecture. Successful completion of all projects results in a 0.3 grade bonus.

Course Outline

1. Introduction
2. Regression: linear regression for prediction and causality, in particular the role of covariates and features; non parametric regression
3. Prediction: LASSO, CART, Random forests, Boosting
4. Causality: Difference-in-Differences, Instrumental Variables, Regression Discontinuity

 

---