Abstract:Learning causal effects from data is a fundamental and well-studied problem across science, especially when the cause-effect relationship is static in nature. However, causal effect is less explored when there are dynamical dependencies, i.e., when dependencies exist between entities across time. Identifying dynamic causal effects from time-series observations is computationally expensive when compared to the static scenario. We demonstrate that the computational complexity of recovering the causation structure for the vector auto-regressive (VAR) model is $O(Tn^3N^2)$, where $n$ is the number of nodes, $T$ is the number of samples, and $N$ is the largest time-lag in the dependency between entities. We report a method, with a reduced complexity of $O(Tn^3 \log N)$, to recover the causation structure to obtain frequency-domain (FD) representations of time-series. Since FFT accumulates all the time dependencies on every frequency, causal inference can be performed efficiently by considering the state variables as random variables at any given frequency. We additionally show that, for systems with interactions that are LTI, do-calculus machinery can be realized in the FD resulting in versions of the classical single-door (with cycles), front and backdoor criteria. We demonstrate, for a large class of problems, graph reconstruction using multivariate Wiener projections results in a significant computational advantage with $O(n)$ complexity over reconstruction algorithms such as the PC algorithm which has $O(n^q)$ complexity, where $q$ is the maximum neighborhood size. This advantage accrues due to some remarkable properties of the phase response of the frequency-dependent Wiener coefficients which is not present in any time-domain approach.

Causal Inference on Process Graphs, Part II: Causal Structure and Effect Identification

Causal Inference on Process Graphs, Part I: The Structural Equation Process Representation

Asymptotic Uncertainty in the Estimation of Frequency Domain Causal Effects for Linear Processes

Causal Structure Recovery of Linear Dynamical Systems: An FFT based Approach

Identifiability and Estimation of Structural Vector Autoregressive Models for Subsampled and Mixed Frequency Time Series

Learning Causal Relations in Multivariate Time Series Data

Confidence in Causal Discovery with Linear Causal Models

Causal Inference From Slowly Varying Nonstationary Processes

Bayesian Inference for Relational Graph in a Causal Vector Autoregressive Time Series

Amortized Causal Discovery: Learning to Infer Causal Graphs from Time-Series Data

A Practical Approach to Causal Inference over Time

Causal inference for process understanding in Earth sciences

Causal discovery and inference: concepts and recent methodological advances

Causal Inference on Time Series using Structural Equation Models

Neural Additive Vector Autoregression Models for Causal Discovery in Time Series

Capturing Causality from Process Data

Identification of Latent Variables From Graphical Model Residuals

Bayesian Vector AutoRegression with Factorised Granger-Causal Graphs

Causal Discovery of Linear Non-Gaussian Causal Models with Unobserved Confounding

Confidence in Causal Inference under Structure Uncertainty in Linear Causal Models with Equal Variances

Graphical criteria for the identification of marginal causal effects in continuous-time survival and event-history analyses