Research map / Probabilistic and scientific modeling / Causal discovery
Cause-Effect Inference in Location-Scale Noise Models: Maximum Likelihood vs. Independence Testing
Xiangyu Sun, Oliver Schulte
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 |
|---|---|---|---|
| Select direction by residual independence instead of likelihood | CAREFL on synthetic LSNM data with Uniform(-1,1) noise and α=10 · LSNM-sine-tanh, Uniform(-1,1), α=10 | accuracy: 0.1 -> 1.0 | Improved |
| Use a misspecified Gaussian prior with likelihood-based selection | CAREFL on synthetic LSNM data with Uniform(-1,1) noise and misleading conditional variances · LSNM-sine-tanh, Uniform(-1,1), α=10 | accuracy: 1.0 -> 0.1 | Hurt |
| Use likelihood-based direction selection when the noise form is correctly specified | CAREFL on synthetic LSNM data · LSNM-sine-tanh, Gaussian(0,1), α=10 | accuracy: 1.0 -> 1.0 | No clear effect |
| Select direction by residual independence | CAREFL on SIM · SIM | accuracy: 0.55 -> 0.80 | Improved |
| Select direction by residual independence | CAREFL on SIM-c · SIM-c | accuracy: 0.58 -> 0.85 | Improved |
| Select direction by residual independence | CAREFL on SIM-ln · SIM-ln | accuracy: 0.84 -> 0.83 | No clear effect |
| Select direction by residual independence | CAREFL on SIM-G · SIM-G | accuracy: 0.82 -> 0.79 | Hurt |
| Select direction by residual independence | CAREFL on Tübingen Cause-Effect Pairs · Tübingen Cause-Effect Pairs | weighted accuracy: 0.73 -> 0.82 | Improved |
| Use a misspecified noise prior with maximum-likelihood selection | CAREFL-M across synthetic LSNM settings · 336 noise-misspecification settings | accuracy: 289 settings (86.01%); 110 settings (32.74%) | Hurt |
| Use residual independence selection rather than likelihood under noise misspecification | CAREFL-H versus CAREFL-M · 336 noise-misspecification settings | accuracy: 333 settings (99.11%); 6 settings (1.79%) | Improved |
| Use residual independence selection with correctly specified noise | CAREFL-H versus CAREFL-M on larger datasets · Synthetic LSNM data, N=5,000 | accuracy: N=5,000 | No clear effect |
Closest papers by meaning
- On the Identifiability and Estimation of Causal Location-Scale Noise Models (ICML 2023 · similarity 0.92)
- Causal Discovery via Transformed Low-Rank Quantile Surfaces (NeurIPS 2026 · similarity 0.89)
- A Skewness-Based Criterion for Addressing Heteroscedastic Noise in Causal Discovery (ICLR 2025 · similarity 0.89)
- Strong and Weak Identifiability of Optimization-based Causal Discovery in Non-linear Additive Noise Models (ICML 2025 · similarity 0.88)
- When Additive Noise Meets Unobserved Mediators: Bivariate Denoising Diffusion for Causal Discovery (NeurIPS 2025 · similarity 0.88)
- Bivariate Causal Discovery using Bayesian Model Selection (ICML 2024 · similarity 0.88)
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.
