Survey reviews conditional independence tests for causal discovery

This survey paper provides a comprehensive review of conditional independence (CI) tests, which are fundamental to constraint-based causal discovery algorithms like PC and FCI. It categorizes common CI methods into six families, including regression, kernel, and machine-learning-based approaches, and examines their assumptions, robustness, and scalability, particularly in high-dimensional and mixed-type biomedical data. The paper also discusses the implications of CI test properties on causal graph recovery and compares the adoption of these methods across R and Python libraries, highlighting open challenges such as mixed-type testing and scalability. AI

IMPACT Provides a foundational review of statistical methods crucial for causal inference in machine learning.

RANK_REASON The item is a survey paper on statistical methods for causal discovery, published on arXiv. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv stat.ML →

• alphaXiv
• arXiv
• CatalyzeX
• DagsHub
• FCI algorithm
• Gotit.pub
• Hugging Face
• Peter-Clark algorithm
• Python
• R
• ScienceCast

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →