Research map / Probabilistic and scientific modeling / Causal discovery
Robust Causal Discovery in Real-World Time Series with Power-Laws
Matteo Tusoni, Giuseppe Masi, Andrea Coletta, Aldo Glielmo, Viviana Arrigoni, Novella Bartolini
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| run Granger test on power-law spectral exponents and amplitudes instead of raw signals | PLaCy · OU non-equilibrium (sigma_m>0), N=5, C=0.5 | F1: 0.39±0.18 -> 0.80±0.17 | Improved |
| use spectral-feature Granger instead of time-domain Granger | PLaCy · OU (sigma_m=0), N=5, C=0.5 | F1: 0.56±0.25 -> 0.73±0.24 | Improved |
| use PLaCy rather than PCMCI | PLaCy · OU synthetic, N=5 | TNR: 0.95±0.05 vs 0.97±0.04 | Mixed |
| use PLaCy instead of frequency-domain baselines | PLaCy · OU synthetic, N=5 | TNR: 0.95±0.05 vs BCGeweke 0.54±0.23 | Improved |
| use PLaCy on real data with non-power-law precipitation | PLaCy · Rivers | F1: 0.47±0.07 -> 0.51±0.10 | Improved |
| use PLaCy instead of PCMCI | PLaCy · AirQuality | F1: 0.25±0.03 -> 0.45±0.04 | Improved |
| use PLaCy instead of PCMCI | PLaCy · AirQuality | TNR: 0.95±0.02 -> 0.66±0.07 | Hurt |
| add amplitude a as covariate in the VAR (ablation) | PLaCy · stationary synthetic data | F1 | Improved |
| add amplitude a as covariate in the VAR (ablation) | PLaCy · non-stationary synthetic data | F1 | No clear effect |
| use amplitude of caused series as input or target (ablation) | PLaCy · non-stationary regimes | false positives | Hurt |
| apply FDR correction (ablation) | PLaCy · synthetic benchmarks | method ranking | No clear effect |
| use slowly varying spectra data | PLaCy · slow-varying spectrum systems | causal discovery quality | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- Rhino — Rhino: Deep Causal Temporal Relationship Learning with History-dependent Noise (ICLR 2023) · better on 4 of 4 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- CausalDynamics: A large‐scale benchmark for structural discovery of dynamical causal models (NeurIPS 2025 · similarity 0.88)
- Flow based approach for Dynamic Temporal Causal models with non-Gaussian or Heteroscedastic Noises (NeurIPS 2025 · similarity 0.88)
- Discovering Mixtures of Structural Causal Models from Time Series Data (ICML 2024 · similarity 0.87)
- CausalTime: Realistically Generated Time-series for Benchmarking of Causal Discovery (ICLR 2024 · similarity 0.87)
- Learning Causal Relations from Subsampled Time Series with Two Time-Slices (ICML 2024 · similarity 0.87)
- CausalRivers - Scaling up benchmarking of causal discovery for real-world time-series (ICLR 2025 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
