MeDIL is a Python package for causal factor analysis, using the measurement dependence inducing latent (MeDIL) causal model framework1. In addition to simple linear Gaussian models, MeDIL also supports deep generative models2.
More information can be found in the documentation.
If you have any questions, suggestions, feedback, or bugs to report, please open an issue on Gitlab or on Github or contact me. Additionally, if you would like to use this package or any of its code in your research, or to contribute to this package, feel free (but not obliged) to contact me.
See LICENSE, which is the GNU Affero General Public License version 3 or later (AGPLv3+).
See CHANGELOG for a history of the already implemented features, works in progress, and future feature ideas.
1. Alex Markham & Moritz Grosse-Wentrup (2020). Measurement Dependence Inducing Latent Causal Models. In Conference on Uncertainty in Artificial Intelligence (UAI) PMLR 124:590–599. URL: http://proceedings.mlr.press/v124/markham20a/markham20a.pdf. arXiv:1910.08778 [stat.ML].
2. Alex Markham, Mingyu Liu, Bryon Aragam, Liam Solus (2023). Neuro-Causal Factor Analysis. prepint. arXiv:2305.19802 [stat.ML].