backShift: Learning causal cyclic graphs from unknown shift interventions

Abstract

We propose a simple method to learn linear causal cyclic models in the presence of latent variables. The method relies on equilibrium data of the model recorded under a specific kind of interventions ("shift interventions"). The location and strength of these interventions do not have to be known and can be estimated from the data. Our method, called backShift, only uses second moments of the data and performs simple joint matrix diagonalization, applied to differences between covariance matrices. We give a sufficient and necessary condition for identifiability of the system, which is fulfilled almost surely under some quite general assumptions if and only if there are at least three distinct experimental settings, one of which can be pure observational data. We demonstrate the performance on some simulated data and applications in flow cytometry and financial time series. The code is made available as R-package backShift.

References in corpus (2)

• Estimating high-dimensional intervention effects from observational data
• Causal Discovery from Changes

Cited by in corpus (12)

• A Survey on Causal Discovery: Theory and Practice
• Causal Discovery in Physical Systems from Videos
• Learning Neural Causal Models from Unknown Interventions
• cuPC: CUDA-based Parallel PC Algorithm for Causal Structure Learning on GPU
• Causal Inference for Time series Analysis: Problems, Methods and Evaluation
• Ancestral Causal Inference
• Robustifying Independent Component Analysis by Adjusting for Group-Wise Stationary Noise
• Permutation-Based Causal Structure Learning with Unknown Intervention Targets
• Identifying Invariant Factors Across Multiple Environments with KL Regression
• Disentangling Observed Causal Effects from Latent Confounders using Method of Moments
• Variance Minimization in the Wasserstein Space for Invariant Causal Prediction
• Matching a Desired Causal State via Shift Interventions