Causal inference

Abstract

Experiments have always been the way to study what the effect is of interventions. Causal inference is an important aspect. In this thesis we gave an introduction to causal inference. We did this by giving an example that illustrates the Fundamental Problem of Causal Inference. The Fundamental Problem of Causal Inference states that it is impossible to observe the values of control and treatment on the same unit and therefore impossible to observe the effect of the treatment on a unit. We used a standard statistical model to later introduce the model for causal inference. The model we used for causal inference is Rubin's model. We assumed that there are two levels of treatment: control and treatment. Both are causes and we determine an effect of a cause always relative to another cause. We discussed a range of assumptions to make it possible to estimate the causal effect. None of them are provable, the best we can do is convince ourselves and others of its correctness. We divided the solutions in two categories: the scientiffic solutions and statistical solutions. The solutions were then used to investigate an issue about alcohol consumption in some newspapers. We concluded that there has to be more awareness about causal inference.