Causality and causal inference from observational data

January 20, 2022, 12:45 - 13:30

Science is concerned with uncovering causal relationships. This can be done with randomized controlled trials. However, in particular in the social sciences, more often than not only observational data is available, as randomized controlled trials are not always possible due to ethical, financial or practical constraints. What is the effect of longterm smoking? We wouldn’t want to force people to try that out. When using observational data, it is difficult to infer causal relationships behind the correlations.

In the first part of this webinar, we will have a look at causal relationships, how they can be expressed in causal networks, and under which conditions direct causal effects can be calculated from observational data. In the second half, we will see mechanisms that are used in practice, such as instrumental variables.

Recordings

• Part 1 - January 20th
• Part 2 - February 3rd (First 6 minutes)
• Part 2 - February 3rd (Remaining ~30 minutes)