Welcome to MeDIL’s documentation!¶
MeDIL is a Python package for causal factor analysis, using the measurement dependence inducing latent (MeDIL) causal model framework [MGW20].
The package is under active development—see the develop branch of the repository on GitLab or its Github mirror.
Features:¶
constraint-based learning of minimum MeDIL causal graphs from marginal independence tests using distance covariance or xi correlation
estimation of causal factor loadings in the linear Gaussian setting or in the nonparametric setting using a variational autoencoder [MLAS23]
sampling from and random generation of linear Gaussian causal factor models
implementation of exact algorithm for minimum edge clique cover (ECC) [GGHuffnerN09] and wrapper for heuristic minimum ECC written in Java [CGM20]
Design principles:¶
scikit-learn style API
basic functionality with minimal dependencies (just NumPy) and optional dependencies for more functionality
as much as possible implemented using NumPy ``ndarray``s and methods for fast performance and wide compatibility
Documentation¶
- Version:
0.7.0
- Date:
Oct 06, 2023