📝 SummarXiv

Abstract

Kaplan--Meier curves are widely used in medical research to evaluate the performance of biomarkers and predict patient outcomes. These curves are often shown without error bands, and even when error bands are provided, they typically only account for the statistical uncertainty resulting from the finite number of patients in the study. In reality, other sources of uncertainty affect the measurements as well. As datasets grow, the statistical uncertainty on the number of patients no longer dominates the overall uncertainty, and other uncertainties are increasingly important to model. The KoMbine package, developed based on procedures used in particle physics, provides the first method to propagate both statistical and systematic uncertainties through the Kaplan--Meier curve estimation processes.