Reproduce “GaussDetect-LiNGAM:Causal Direction Identification without Gaussianity test” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Ziyi Ding, Xiao-Ping Zhang

Researchers at Tsinghua University developed GaussDetect-LiNGAM for bivariate causal discovery, which bypasses the need for direct Gaussianity tests of noise terms by demonstrating their equivalence to independence tests between regressors and residuals. This approach enhances the reliability and computational efficiency of causal direction identification within the LiNGAM framework.

Key insights

  • A fundamental theoretical equivalence exists between testing noise Gaussianity in one causal direction and testing independence between the predictor and residual in the reverse regression model.
  • Robust kernel-based independence tests can reliably serve as proxies for assessing the effective Gaussianity of noise, thereby making LiNGAM more dependable.
  • The proposed method automatically discerns whether the noise environment permits LiNGAM identifiability, leading to more transparent and responsible causal inference.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.