TWO-SAMPLE TEST FOR RANDOMLY CENSORED DATA

Abstract

In this paper, a nonparametric test has been proposed for the two-sample scale problem, when sample observations are randomly right censored. The proposed test is based on the extremes of observations as an extension of commonly used Gehan’s test for two-sample problem. Critical values are obtained through simulation for various lifetime distributions at different sample sizes. Power performance for the proposed test is studied considering various distributions. On comparing with the Gehan’s test, it is found that the proposed test has more statistical power and efficiency for some special cases. An illustration with real-life data set is also provided.