Computing the nonparametric estimator of the survivor function when all observations are either left- or right-censored with tied observation times

A data set of missiles tested at various times consists entirely of left- and right-censored observations. We present an algorithm for computing the nonparametric maximum likelihood estimator of the survivor function. When there are a significant number of tied observations, the algorithm saves significant computation time over a direct implementation of the survivor function estimate given in Andersen and Rønn, (Biometrics 1995; 51:323–9).