Doubly Robust Conformalized Survival Analysis with Right-Censored Data

2024-12-12

arXiv.org

AI Summary

To address the lack of finite-sample statistical guarantees and susceptibility to model misspecification in predicting lower bounds for survival times under right-censored data, this paper proposes a doubly robust conformal survival inference framework. The method integrates machine learning–driven imputation of censoring times, weighted conformal inference, and semiparametric survival modeling. It is the first to deliver prediction intervals for survival time lower bounds with guaranteed finite-sample coverage under arbitrary censoring mechanisms—not restricted to Type-I censoring. We establish theoretical double robustness: coverage is maintained if either the survival or the censoring model is correctly specified; empirically, the method exhibits strong robustness to model misspecification. In extensive simulations and real-data experiments, our approach achieves significantly narrower (i.e., more informative) prediction intervals than existing methods while strictly maintaining the nominal coverage level—even under severe survival model misspecification.

Technology Category

• Machine Learning: Calibration & Uncertainty Quantification
• Reasoning under Uncertainty: Causality
• Cognitive Modeling & Cognitive Systems: Conceptual Inference and Reasoning

Application Category

• User Modeling, Personalization and Recommendation: User privacy protection in personalized systems
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Responsible Web: Measurement, analysis, and circumvention of Web censorship

Abstract

We present a conformal inference method for constructing lower prediction bounds for survival times from right-censored data, extending recent approaches designed for more restrictive type-I censoring scenarios. The proposed method imputes unobserved censoring times using a machine learning model, and then analyzes the imputed data using a survival model calibrated via weighted conformal inference. This approach is theoretically supported by an asymptotic double robustness property. Empirical studies on simulated and real data demonstrate that our method leads to relatively informative predictive inferences and is especially robust in challenging settings where the survival model may be inaccurate.

Problem

• Extends conformal inference for right-censored survival data
• Uses machine learning to impute unobserved censoring times
• Ensures robustness with inaccurate survival models

Innovation

• Conformal inference for survival time prediction
• Machine learning imputes censoring times
• Double robustness with weighted calibration