Two Sample Test for Censored Data Based on Sub-Sample Medians

Abstract

In hypothesis testing, the purpose of two-sample problem is to determine the statistical significance of the difference between two independent populations. This paper proposes a nonparametric test statistic based on U -statistic, useful for two-sample scale problem when dealing with randomly right censored data. Proposed test concentrates on comparing the medians of sub-samples from two populations. A simulation study is given for critical values and power of the proposed test statistic at various sample sizes and considering different lifetime distributions. Asymptotic relative efficiencies show that the proposed test is more efficient in comparison to some existing tests available in literature. A real-life data application is also given.