TILS1500 Causal models (2–5 cr)

Causality is a central concept in science. In causal inference, conclusions are made about how one factor (cause) affects another (effect). Experimental research is closely related to causal inference: in an experiment, a factor is changed or varied in a controlled manner and the change in response is observed. Causal inference is also possible in observational research if the assumptions made about confounding factors are valid. Therefore, in causal inference, conclusions are influenced not only by observations but also by assumptions about causal relations between variables.

The course deals with the modeling of causal relations, causal inference, and the estimation of causal effects. The most important tool used is the functional causal model introduced by Judea Pearl.

A student who has completed the course

• knows the concept of causality,
• can represent causal relations with graphical models,
• knows Rubin’s and Pearl’s causal models and related terminology,
• can use software for causal inference,
• can make conclusions about causal effects,
• knows the connection of causal models to structural equation models,
• can estimate causal effects from observational data,
• can critically evaluate claims about cause-effect relationships presented in the media.

In addition, a student who has completed the course for 5 credits:

• masters the theory related to the topics discussed,
• can apply do-calculus,
• understands the operating principles of some algorithms used in causal reasoning,
• can make counterfactual conclusions.

For students of statistics and data science, the scope of the course is 5 credits. Other students can choose the scope of the course to be 2 credits or 5 credits.

The course is held every one and a half years.

5 credits: Statistical inference 1, Generalized linear models 1, basic skills with R software

2 credits: From data to model, basic skills with R software