Telling cause from effect based on high-dimensional observations

Abstract

We describe a method for inferring linear causal relations among multi-dimensional variables. The idea is to use an asymmetry between the distributions of cause and effect that occurs if both the covariance matrix of the cause and the structure matrix mapping cause to the effect are independently chosen. The method works for both stochastic and deterministic causal relations, provided that the dimensionality is sufficiently high (in some experiments, 5 was enough). It is applicable to Gaussian as well as non-Gaussian data.

13 pages, 5 figures

References in corpus (2)

• On the Identifiability of the Post-Nonlinear Causal Model
• Causal inference using the algorithmic Markov condition

Cited by in corpus (11)

• Distinguishing cause from effect using observational data: methods and benchmarks
• Causality for Machine Learning
• Inferring deterministic causal relations
• Testing whether linear equations are causal: A free probability theory approach
• Telling cause from effect in deterministic linear dynamical systems
• Latent Instrumental Variables as Priors in Causal Inference based on Independence of Cause and Mechanism
• Predicting with High Correlation Features
• A Kernel Embedding-based Approach for Nonstationary Causal Model Inference
• The Effect of Noise Level on Causal Identification with Additive Noise Models
• Detecting confounding in multivariate linear models via spectral analysis
• Cause-effect inference through spectral independence in linear dynamical systems: theoretical foundations