Functional Clustering of Survival Data via Smoothed Log-Hazard Trajectories: A Risk-Dynamics Perspective
Researchers have developed a new method for clustering survival data by analyzing the instantaneous risk, or hazard function, over time. This approach models smoothed log-hazard trajectories as functional objects and uses Functional Principal Component Analysis for clustering. The methodology was tested through simulations and applied to clinical datasets, demonstrating its ability to provide interpretable representations of temporal risk dynamics. AI
IMPACT Introduces a novel statistical framework for analyzing complex temporal risk dynamics in survival data.