LoRSI

LoRSI computes a stability interval for log-rank test P-values to quantify uncertainty introduced by sample labeling errors in survival analysis.

Key Features:

• Addressing Labeling Errors: Quantifies the effect of sample labeling errors on group assignments used by the log-rank test.
• Stability Interval Calculation: Computes a stability interval around the original log-rank P-value that reflects uncertainty due to labeling inaccuracies.
• Algorithmic Approach: Employs a novel, efficiently computable algorithm to derive the stability interval with a rigorous correctness proof.
• Python Implementation: Provided as a Python implementation for integration into computational analysis workflows.
• Demonstration on Datasets: Validated on multiple datasets to demonstrate behavior and performance in real-world survival analyses.

Scientific Applications:

• Oncology: Assesses robustness of survival differences and treatment effect conclusions in oncology studies using time-to-event endpoints.
• Epidemiology: Evaluates sensitivity of survival analyses in epidemiological studies to sample labeling errors.
• Biomedical time-to-event research: Provides uncertainty quantification for log-rank P-values to inform conclusions about treatment effects or disease progression in biomedical studies involving time-to-event data.

Methodology:

Defines uncertainty from sample labeling errors in the context of the log-rank test and computes a stability interval for the log-rank P-value using a provably correct, efficiently computable algorithm.