Distinguishing cause from effect using observational data: methods and benchmarks

Abstract

The discovery of causal relationships from purely observational data is a fundamental problem in science. The most elementary form of such a causal discovery problem is to decide whether X causes Y or, alternatively, Y causes X, given joint observations of two variables X, Y. An example is to decide whether altitude causes temperature, or vice versa, given only joint measurements of both variables. Even under the simplifying assumptions of no confounding, no feedback loops, and no selection bias, such bivariate causal discovery problems are challenging. Nevertheless, several approaches for addressing those problems have been proposed in recent years. We review two families of such methods: Additive Noise Methods (ANM) and Information Geometric Causal Inference (IGCI). We present the benchmark CauseEffectPairs that consists of data for 100 different cause-effect pairs selected from 37 datasets from various domains (e.g., meteorology, biology, medicine, engineering, economy, etc.) and motivate our decisions regarding the "ground truth" causal directions of all pairs. We evaluate the performance of several bivariate causal discovery methods on these real-world benchmark data and in addition on artificially simulated data. Our empirical results on real-world data indicate that certain methods are indeed able to distinguish cause from effect using only purely observational data, although more benchmark data would be needed to obtain statistically significant conclusions. One of the best performing methods overall is the additive-noise method originally proposed by Hoyer et al. (2009), which obtains an accuracy of 63+-10 % and an AUC of 0.74+-0.05 on the real-world benchmark. As the main theoretical contribution of this work we prove the consistency of that method.

101 pages, second revision submitted to Journal of Machine Learning Research

References in corpus (9)

• The Solar Cycle
• On the Identifiability of the Post-Nonlinear Causal Model
• On Causal and Anticausal Learning
• Inferring deterministic causal relations
• Information Theoretical Estimators Toolbox
• Gaussian Process Networks
• Telling cause from effect based on high-dimensional observations
• Testing whether linear equations are causal: A free probability theory approach
• A simpler condition for consistency of a kernel independence test

Cited by in corpus (99)

• Kernel Mean Embedding of Distributions: A Review and Beyond
• Learning Decentralized Controllers for Robot Swarms with Graph Neural Networks
• Inferring causal structure: a quantum advantage
• Estimating individual treatment effect: generalization bounds and algorithms
• Causality for Machine Learning
• A Survey on Causal Discovery: Theory and Practice
• Foundations of Structural Causal Models with Cycles and Latent Variables
• Limits to causal inference with state-space reconstruction for infectious disease
• Experimental Test of Nonlocal Causality
• Towards a Learning Theory of Cause-Effect Inference
• Analysis of cause-effect inference by comparing regression errors
• Telling Cause from Effect using MDL-based Local and Global Regression
• CoPhy: Counterfactual Learning of Physical Dynamics
• Learning Directed Acyclic Graphs with Penalized Neighbourhood Regression
• Causal Inference via Kernel Deviance Measures
• Structural Agnostic Modeling: Adversarial Learning of Causal Graphs
• On the Optimality of Gaussian Kernel Based Nonparametric Tests against Smooth Alternatives
• An Evaluation Toolkit to Guide Model Selection and Cohort Definition in Causal Inference
• Learning Sparse Nonparametric DAGs
• Causal Generative Neural Networks
• Identifying Best Interventions through Online Importance Sampling
• Causal Discovery from Incomplete Data: A Deep Learning Approach
• Bayesian Learning of Kernel Embeddings
• Causal Inference for Time series Analysis: Problems, Methods and Evaluation
• Robust inference of causality in high-dimensional dynamical processes from the Information Imbalance of distance ranks
• Neuropathic Pain Diagnosis Simulator for Causal Discovery Algorithm Evaluation
• Boosting Synthetic Data Generation with Effective Nonlinear Causal Discovery
• MONK -- Outlier-Robust Mean Embedding Estimation by Median-of-Means
• The Randomized Causation Coefficient
• Causal Discovery using Compression-Complexity Measures
• Invariant Representation Learning for Treatment Effect Estimation
• Causal, Bayesian, & Non-parametric Modeling of the SARS-CoV-2 Viral Load Distribution vs. Patient's Age
• Detecting non-causal artifacts in multivariate linear regression models
• Causal Discovery Using Proxy Variables
• Causality and Batch Reinforcement Learning: Complementary Approaches To Planning In Unknown Domains
• The Arrow of Time in Multivariate Time Series
• Causality in extremes of time series
• Probabilistic Active Learning of Functions in Structural Causal Models
• Switching Regression Models and Causal Inference in the Presence of Discrete Latent Variables
• Distinguishing Cause from Effect Using Quantiles: Bivariate Quantile Causal Discovery
• Estimating Causal Direction and Confounding of Two Discrete Variables
• Latent Instrumental Variables as Priors in Causal Inference based on Independence of Cause and Mechanism
• A polynomial-time algorithm for learning nonparametric causal graphs
• Causal Inference by Stochastic Complexity
• A Critical View of the Structural Causal Model
• Vision Paper: Causal Inference for Interpretable and Robust Machine Learning in Mobility Analysis
• TimeGraph: Synthetic Benchmark Datasets for Robust Time-Series Causal Discovery
• On Data-Driven Computation of Information Transfer for Causal Inference in Dynamical Systems
• A Bayesian Model for Bivariate Causal Inference
• From controlled to undisciplined data: estimating causal effects in the era of data science using a potential outcome framework
• Leveraging directed causal discovery to detect latent common causes
• Unsupervised Machine Learning Discovery of Chemical and Physical Transformation Pathways from Imaging Data
• Causal Inference via Conditional Kolmogorov Complexity using MDL Binning
• Causal Mosaic: Cause-Effect Inference via Nonlinear ICA and Ensemble Method
• Optimal experimental design via Bayesian optimization: active causal structure learning for Gaussian process networks
• Exclusivity graph approach to Instrumental inequalities
• Causal World Models by Unsupervised Deconfounding of Physical Dynamics
• Causal inference using deep neural networks
• Confidence in Causal Discovery with Linear Causal Models
• Identifying Causal Structure in Dynamical Systems
• Analysis of Drifting Features
• Merging joint distributions via causal model classes with low VC dimension
• Bayesian Allocation Model: Inference by Sequential Monte Carlo for Nonnegative Tensor Factorizations and Topic Models using Polya Urns
• Optimal Experiment Design for Causal Discovery from Fixed Number of Experiments
• Reparametrization Invariance in non-parametric Causal Discovery
• Quantum causal inference in the presence of hidden common causes: An entropic approach
• Causal Structure Learning: a Bayesian approach based on random graphs
• The Effect of Noise Level on Causal Identification with Additive Noise Models
• Goodness-of-Fit Tests for Linear Non-Gaussian Structural Equation Models
• Dynamical systems theory for causal inference with application to synthetic control methods
• Optimal transport for causal discovery
• Analysis via Orthonormal Systems in Reproducing Kernel Hilbert $C^{\ast }$-Modules and Applications
• Independent mechanism analysis, a new concept?
• Causal Autoregressive Flows
• Conditionally-additive-noise Models for Structure Learning
• Quantifying intrinsic causal contributions via structure preserving interventions
• Identifiability of AMP chain graph models
• Unifying DAGs and UGs
• Causal Inference on Multivariate and Mixed-Type Data
• Ancestral causal learning in high dimensions with a human genome-wide application
• The Crossover Process: Learnability and Data Protection from Inference Attacks
• Sensitivity Maps of the Hilbert-Schmidt Independence Criterion
• Estimation of Bivariate Structural Causal Models by Variational Gaussian Process Regression Under Likelihoods Parametrised by Normalising Flows
• Causal inference for data-driven debugging and decision making in cloud computing
• A Causal Direction Test for Heterogeneous Populations
• Information-Theoretic Approximation to Causal Models
• Causal network learning with non-invertible functional relationships
• Understanding Human Judgments of Causality
• Adversarial Orthogonal Regression: Two non-Linear Regressions for Causal Inference
• Causal Discovery by Kernel Intrinsic Invariance Measure
• Parallel ensemble methods for causal direction inference
• Autoregressive flow-based causal discovery and inference
• Encoding Causal Macrovariables
• Efficiently Disentangle Causal Representations
• Warped Gaussian Processes in Remote Sensing Parameter Estimation and Causal Inference
• Identifying Causal Influences on Publication Trends and Behavior: A Case Study of the Computational Linguistics Community
• Causal Inference in Geoscience and Remote Sensing from Observational Data
• Causal Inference in Geosciences with Kernel Sensitivity Maps
• Causal Identification with Additive Noise Models: Quantifying the Effect of Noise