Answers are grounded in this paper and cite their sources.
A series of recent papers investigate the two implementations of the Schoenfeld-based test for diagnosing proportional hazard violations in Cox duration models. These studies show surprising performance differences between the implementations—which one we use can affect our conclusions about whether a covariate violates the proportional hazards assumption. Importantly, though, these studies examine situations in which we would generally expect no performance differences between the implementations. How do the different implementations perform in situations where we would expect differences? I use Monte Carlo simulations to investigate.
No takes yet. Share an insight, caveat, or question.
Shawna K. Metzger (2024) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: