![IconGems-Print](images/IconGems-Print.png)
167 – Special Issues in Modeling
On Effect Sizes for Nonparametric Comparison of Censored Survival Outcomes
Yongzhao Shao
New York University School of Medicine
Zhaoyin Zhu
NYU
Survival outcomes are frequently randomly censored with unknown censoring distributions. Due to complexities caused by censoring, useful effect sizes for nonparametric comparison of censored survival outcomes have not been systematically investigated despite existence of several well known nonparametric tests such as the log-rank test and Wilcoxon test. Effect size generally emphasizes the magnitude of the difference between the studied endpoints rather than confounding this with sample size as in the case of p-value. This paper investigates weakness and advantages of existing and newly proposed effect sizes for the nonparametric comparison of time-to-event outcomes.