A non-parametric test for interval-censored failure time data with application to AIDS studies

Interest often centres on the comparison of failure time distributions based on interval-censored failure time data such as in the work by Finkelstein, in which she proposed a score test under continuous proportional hazards model. In this article, we consider a different situation in which the underlying failure time is a discrete variable or the observed times correspond with a discrete scale. To compare failure time distributions in these situations, we propose a non-parametric test, a generalization of the usual logrank test for right-censored failure time data. Simulation results indicate that the test performs satisfactorily.