A novel approach to simulating duration data addresses a key limitation of the standard Cox proportional hazards model. Unlike previous methods, this technique generates durations directly from the model's inherent data generating process rather than relying on parametric survival distributions.
The method involves fitting a cubic spline baseline hazard function by randomly drawing points for each subject in the dataset. This ensures greater flexibility and aligns more closely with when researchers choose to use the Cox model over parametric alternatives.
This improved simulation strategy allows political scientists to test hypotheses under conditions that better reflect their research goals, leading to findings potentially applicable across diverse contexts without requiring specific distributional assumptions.
Key Improvements:
Directly simulates durations from the inherent* data generating process of the Cox model.
| Simulating Duration Data for the Cox Model was authored by Jeffrey J. Harden and Jonathan Kropko. It was published by Cambridge in PSR&M in 2019. |
