FLEXI-Haz: Semiparametric Deep Survival Inference Beyond Proportional Hazards
FLEXI-Haz introduces a deep neural network model to analyze survival data beyond simple proportional hazards assumptions. It combines a linear component for interpretable treatment effects with a flexible deep learning component to capture complex, time-dependent interactions among nuisance variables. This allows researchers to estimate risks accurately while maintaining the ability to interpret how specific factors influence survival over time.