CEDAR: Causal Edge Discovery for Autoregressive Processes

2026-07-22 · Source: Machine Learning · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Data Science & Analytics · Depth: Expert, quick

Summary

CEDAR, a constraint-based method, is proposed for lagged causal edge discovery in sparse autoregressive time series. This technique screens candidate cross-variable lags using AR(1)-residualized, U-centered distance correlation. It then applies two targeted conditional-independence tests per significant candidate, accepting at most one lag per ordered pair. A stable MCI pruning step removes indirect edges, and optional deterministic C-nodes can adjust for trend-like nonstationarity. CEDAR is particularly effective with scarce data and variables exhibiting lag-1 self-dynamics, requiring O(d^2) CI tests after screening in sparse regimes while maintaining edge-level interpretability. However, methods with richer conditioning sets become preferable as time series length T grows or when higher-order autoregressive or simultaneous multi-lag effects are common.

Key takeaway

For research scientists analyzing sparse autoregressive time series with limited data, CEDAR offers an efficient, interpretable method for lagged causal edge discovery. If your datasets are characterized by scarce observations and variables primarily exhibiting lag-1 self-dynamics, you should consider implementing CEDAR to identify causal relationships. Be aware that for longer time series or complex multi-lag effects, alternative methods with richer conditioning sets may be more suitable.

Key insights

CEDAR discovers lagged causal edges in sparse autoregressive time series, optimized for scarce data and lag-1 self-dynamics.

Principles

• CEDAR excels with scarce data and lag-1 self-dynamics.
• Richer conditioning sets are better for longer series or higher-order effects.

Method

CEDAR screens lags via AR(1)-residualized distance correlation, applies two conditional-independence tests per candidate, accepts one lag per pair, then prunes indirect edges with MCI, optionally adjusting for nonstationarity.

In practice

• Use CEDAR for sparse time series with limited data.
• Prioritize CEDAR when variables exhibit strong lag-1 self-dynamics.

Topics

• Causal Discovery
• Autoregressive Models
• Time Series Analysis
• Conditional Independence Testing
• Lagged Effects
• Sparse Data

Best for: AI Scientist, Research Scientist

Related on AIssential

• FoundCause: Causal Discovery with Latent Confounders from Observational Data — FoundCause uses amortized learning and specific architectural biases to discover causal graphs and latent confounders from observational data.
• Beyond Additivity: Causal Discovery in Location-Scale Noise Models with Hidden Variables — Causal discovery in location-scale noise models with hidden variables is identifiable, even when causes affect both mean and variance.
• Unsupervised Causal Abstractions Discovery — This work proposes an unsupervised method to discover high-level causal abstractions directly from low-level observational data.
• Estimating Mutual Information between Time Series and Temporal Event Sequences Across Diverse Analysis Tasks — A nonparametric mutual information estimator quantifies dependence between heterogeneous time series and event sequences without transformation or discretization.
• Beyond Means: Topological Causal Effects under Persistent-Homology Ignorability — Persistent homology enables identifying causal effects on outcome distribution shapes, addressing limitations of mean-based estimands.
• Moment Matters: Mean and Variance Causal Graph Discovery from Heteroscedastic Observational Data — Separating mean and variance causal influences improves interpretability and intervention design in heteroscedastic systems.

See Counsel's argued verdicts on the open AI decisions leaders are weighing →

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Machine Learning.