Evaluation of Competing Risks Methods

• In Event: 35.034 - Some Sharp Statistics

Abstract

A simulation study is conducted to evaluate common methods of analyzing competing risks data. Preliminary results appear to be closely related to current literature on previously tested condition combinations. Separate modeling via Cox proportional hazards regression appears to be the most consistent at matching set failure rate parameters for these conditions. When investigating data with differing event time distributions however, preliminary results suggest separate modeling via Cox proportional hazards regression is inconsistent and outperformed by other models. Detailed results for all sets of conditions are forthcoming in the full paper.

Author

• Brian Rickard, University of Arkansas