to confirm with Stats / MCO regarding ties methods in `get_cox_pairwise_df()`

[jszczypinski]

Statistical Alignment: Confirming Tie-Handling Mapping Between R and SAS

We need Stats / MCO team to confirmthe alignment of tie-handling methods between R (survival::coxph) and SAS (PROC PHREG).

The Mapping Confirmation:

• R ties = "exact" is the identical mathematical equivalent of SAS TIES = DISCRETE. Both execute the exact discrete partial likelihood (conditional logistic model).
• SAS TIES = EXACT calculates the exact marginal likelihood (based on Kalbfleisch & Prentice), which is not implemented in R's survival package. Do we have to implement it?

References for the Stats / MCO:
These references explicitly confirm the method mapping, demonstrating that SAS TIES=DISCRETE is the identical mathematical equivalent to R's ties="exact" in Cox regression models:

1. UC Davis Lecture Notes (EPI 204): Lecture 11: Survival Regression (PDF)
2. R-Help Archives: Official Mailing List Discussion on Coxph Tie Handling
3. PSI AIMS CAMIS Project: R and SAS Survival Software Comparison
4. R for SAS Users Guide: Survival Analysis Syntax and Mapping Reference

[jszczypinski]

@shajoezhu here is the issue regarding ties methods