The standard Schoenfeld residual test for proportional hazards models is often misapplied due to arbitrary time scale choices. This article demonstrates that such decisions significantly impact research conclusions, revealing a gap between statistical practice and data realities.
New Insight Needed:
* Relying solely on the Schoenfeld test without considering specific transformations or their justification leads to potentially misleading results.
* The choice of time transformation profoundly affects model validity in political science event history studies.
Data Matters Most:
* Ignoring outlier survival times and censoring levels exacerbates misinterpretation risks.
* Simulations show that the seemingly minor decision about time scales can dramatically alter findings.
Instead of just adding a test, scholars should integrate exploratory data analysis into their modeling process. This approach provides a more robust framework for evaluating proportional hazards assumptions in event studies.
| Reassessing Schoenfeld Residual Tests of Proportional Hazards in Political Science Event History Analyses was authored by Sunhee Park and David Hendry. It was published by Wiley in AJPS in 2015. |
