This is a repository to store notebooks of statistical analysis in graphical modelling and causal inference (MaRDI Project, TA3)
Here we provide the short description of the available notebooks and their possible usage.
Notebook_01 - introduction to the problem of graphical modelling and causal inference. Implementation of the algorithms available in the package bnlearn on the dataset alarm which were collected from Zenodo repository.
Notebook_02 - introduction to the problem of subgraph (subset) selection in a "large" graph with available ground truth such that the independence relationships in the selected graph do not contradict to the ones in the supergraph. Implementation of different heuristics based on alarm graph and their benchmarking.
Notebook_03 - based on the data collected by telekom company, estimate which covariate cause clients to become churn. Also a comparison to known methods which provide feature importance is presented. This notebook can also serve as an introductory notebook in the courses of statistical analysis.
Notebook_04 - benchmarking datasets from Tuebingen cause-effect repository by means of a statistical test which uses kernel-based HSIC methodology. Work in progress.
Notebook_05 - a notebook which introduces the problem of DAG estimation as a non-convex optimization problem over euclidean space. Implementation uses dataset from Bayesys collection collected and published earlier in zenodo community. Work in progress (extension to non-linear optimization problem).