# Practical Bounding of Counterfactual Inferences by Credal Networks

Consider an observational (or interventional or hybrid) dataset.
Say that you are interested in causal inference and in particular in a counterfactual analysis.
You can use the dataset based on recovery/treatment/gender data, but if you have your own data is even better.
I can support you during the project.
The main steps are:

- Identification of the causal, counterfactual, query we want to answer.
- Identification of the underlying causal graph and possible latent confounders.
- Specification (expert-based or canonical) of the structural equations.
- Implementation of the equivalent credal network.
- Computation of the bounds and analysis of the results.

Even if we have dedicated software tools for that, for small models like the one proposed to
the participants, the analysis can also be sketched on paper (or in a Python notebook).