A non-parametric comparison of two samples one of which is censored