This is a group project, done as part of a Machine Learning in Physical World module at Uniersity of Cambridge. Our group decided to explore topic of causal inference, especially what are the difficulties in answering counterfactual questions wiht observational data. We found this topic interesting, due to its usefulness when evaluating isolated effects of actions which we took in the past, for example, whta was the effectiveness of different non-pharmaceutical interventions (NPIs, i.e. school closure, 2m social distancing order etc.) introduced by governments in fighting the COVID pandemic? Governments can utilise data from across different countries, however the inference is not straightforwrd because there might be some confounding effects which will affect both the treatment and outcome variables. \
In this project we explore how probabilisitic modelling can be applied to answering counterfactual questions.