A NOVEL CENSORING RANK TEST FOR SURVIVAL DATA

Abstract

The objective of this research was to overcome the problem of moderate to heavy censoring in survival analysis for comparing the survival functions of two groups. In literature several tests are available, e.g. the unweighted log-rank test and the weighted Wilcoxon, tarone ware tests. These tests considered only the events and give no importance to the censoring. But in real life some times, the data set consists of more censored observations than the events. To solve this issue a novel censoring rank test is developed in this research. The test considered the three quartiles i.e. Q1, Q2 and Q3 respectively. Performance of the new method with the existing methods compared by the simulation study. Results of the study showed that the novel test perform much better than the tests available in the literature.The proposed methods along with the available methods are applied to two real data sets i.e. Leukemia and Thalassaemia data sets. Results of the applications are also support the novel test.