A. Lago Balseiro, J. de Uña Álvarez, J. C. Pardo-Fernández
In this talk we introduce a Kolmogorov-Smirnov statistic for the two-sample problem under random left truncation. A simulation study is carried out to analyse its distribution under the null hypothesis. The results show the test is not distribution-free, in contrast to the Kolmogorov-Smirnov test for complete data. This leads to propose a bootstrap resampling plan, which allows to approximate the distribution of the proposed statistic test under the null hypothesis. We compare the proposed Kolmogorov-Smirnov test to the log-rank test under different situations: proportional hazards, where the log-rank test is optimal, and nonproportional hazards. Scenarios under the alternative hypothesis for which the log-rank test has no power whereas the Kolmogorov-Smirnov test does are also shown. Lastly, a real dataset is analysed.
Keywords: bootstrap, hypothesis testing, left truncation, survival analysis
Scheduled
GT18 Non-parametric statistics I. Non-parametric hypothesis tests
June 9, 2022 10:10 AM
A15