Testing the equality among distribution functions from independent and right censored samples via Cramér--von Mises criterion

The traditional Cramér--von Mises criterion is used in order to develop a test to compare the equality of the underlying lifetime distributions in the presence of independent censoring times. Its asymptotic distribution is proved and a resampling plan, which is valid for unbalanced data situations, is proposed. Its statistical power is studied and compared with commonly used linear rank tests by Monte Carlo simulations and a real data analysis is also considered. It is observed that the new test is clearly more powerful than the traditional ones when there exists <italic>no uniform dominance</italic> among involved distributions and in the presence of late differences. Its statistical power is also good in the other considered scenarios.