New research questions survival model evaluation methods

A new paper published on arXiv explores the properness of scoring rules in survival model evaluation, particularly under censoring. The research, led by Raphael Sonabend, introduces a concept of marginal properness and demonstrates that commonly used rules like SBS and ISBS can become improper with finite follow-up or cure fractions. The study highlights how these theoretical issues can lead to misleading comparisons between models, emphasizing the need for improved evaluation methods in survival analysis. AI

IMPACT Highlights potential flaws in evaluating AI models used in survival analysis, impacting the reliability of model comparisons.

RANK_REASON Academic paper on statistical theory for model evaluation. [lever_c_demoted from research: ic=1 ai=0.7]

Read on arXiv cs.LG →

• arXiv
• automated machine learning
• Raphael Sonabend

• paper
• other

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