New framework enhances causal inference with reliable uncertainty quantification

Researchers have developed a new framework called Deconditional Gaussian Process (DGP) to improve causal inference methods, specifically instrumental variable (IV) and proximal causal learning (Proxy). This framework addresses the limitation of existing methods by providing reliable epistemic uncertainty quantification. The DGP framework integrates popular kernel estimators and offers principled, well-calibrated uncertainty through posterior variance, enabling systematic model selection via marginal log-likelihood optimization. Empirical results show strong predictive performance and informative uncertainty quantification. AI

RANK_REASON This is a research paper published on arXiv detailing a new statistical framework for causal inference. [lever_c_demoted from research: ic=1 ai=0.7]

Read on arXiv stat.ML →

• arXiv
• Deconditional Gaussian Process
• Gaussian Processes
• proximal causal learning
• Zhang Yuqi

• paper
• other

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