Quantifying causal influences

Abstract

Many methods for causal inference generate directed acyclic graphs (DAGs) that formalize causal relations between $n$ variables. Given the joint distribution on all these variables, the DAG contains all information about how intervening on one variable changes the distribution of the other $n-1$ variables. However, quantifying the causal influence of one variable on another one remains a nontrivial question. Here we propose a set of natural, intuitive postulates that a measure of causal strength should satisfy. We then introduce a communication scenario, where edges in a DAG play the role of channels that can be locally corrupted by interventions. Causal strength is then the relative entropy distance between the old and the new distribution. Many other measures of causal strength have been proposed, including average causal effect, transfer entropy, directed information, and information flow. We explain how they fail to satisfy the postulates on simple DAGs of $\le 3$ nodes. Finally, we investigate the behavior of our measure on time-series, supporting our claims with experiments on simulated data.

Published in at http://dx.doi.org/10.1214/13-AOS1145 the Annals of Statistics (http://www.imstat.org/aos/) by the Institute of Mathematical Statistics (http://www.imstat.org)

References in corpus (4)

• On the Identifiability of the Post-Nonlinear Causal Model
• Information-theoretic approach to the study of control systems
• Causal Inference on Discrete Data using Additive Noise Models
• Identifiability of Causal Graphs using Functional Models

Cited by in corpus (16)

• Detecting causal associations in large nonlinear time series datasets
• Quantum stochastic processes and quantum non-Markovian phenomena
• When the map is better than the territory
• Quantifying information transfer and mediation along causal pathways in complex systems
• Quantum Inflation: A General Approach to Quantum Causal Compatibility
• Experimental Test of Nonlocal Causality
• Classical causal models for Bell and Kochen-Specker inequality violations require fine-tuning
• Entropic nonsignaling correlations
• Information Flow in Computational Systems
• Measuring the integrated information of a quantum mechanism
• Direct and Indirect Effects -- An Information Theoretic Perspective
• Can Transfer Entropy Infer Information Flow in Neuronal Circuits for Cognitive Processing?
• Monogamy of Temporal Correlations: Witnessing non-Markovianity Beyond Data Processing
• Quantifying Quantum Causal Influences
• An axiomatic measure of one-way quantum information
• Fluctuation-response theorem for Kullback-Leibler divergences to quantify causation