Learning Causal or Non-Causal Graphical Models Using Information
Theory
Description
Multivariate Information-based Inductive Causation, better known
by its acronym MIIC, is a causal discovery method, based on information
theory principles, which learns a large class of causal or non-causal
graphical models from purely observational data, while including the effects
of unobserved latent variables. Starting from a complete graph, the method
iteratively removes dispensable edges, by uncovering significant information
contributions from indirect paths, and assesses edge-specific confidences
from randomization of available data. The remaining edges are then oriented
based on the signature of causality in observational data. The recent more
interpretable MIIC extension (iMIIC) further distinguishes genuine causes
from putative and latent causal effects, while scaling to very large
datasets (hundreds of thousands of samples). Since the version 2.0, MIIC
also includes a temporal mode (tMIIC) to learn temporal causal graphs from
stationary time series data. MIIC has been applied to a wide range of
biological and biomedical data, such as single cell gene expression data,
genomic alterations in tumors, live-cell time-lapse imaging data
(CausalXtract), as well as medical records of patients. MIIC brings unique
insights based on causal interpretation and could be used in a broad range
of other data science domains (technology, climatology, economy, ...).
For more information, you can refer to:
Simon et al., eLife 2024, ,
Ribeiro-Dantas et al., iScience 2024, ,
Cabeli et al., NeurIPS 2021, ,
Cabeli et al., Comput. Biol. 2020, ,
Li et al., NeurIPS 2019, ,
Verny et al., PLoS Comput. Biol. 2017, ,
Affeldt et al., UAI 2015, .
Changes from the previous 1.5.3 release on CRAN are available at
.