Abstract
A k-sample median test for censored data is proposed. The test is more sensitive to differences in median survival times than to differences in shape of the survival curves. The asymptotic distribution of the test statistic is derived under both the null and the alternative hypotheses and its variance is shown to take a simple form. The median test is compared to several other tests for censored data by computing asymptotic relative efficiencies and performing a Monte Carlo simulation under various conditions. The procedure described is applied to data from a colorectal cancer clinical trial.
Key Words: