A concordance test for independence in the presence of censoring.

An extension of the definition of Kendall's coefficient of concordance is proposed for testing independence in bivariate censored survival times. The test statistic counts the number of known concordances minus the number of known discordances. The variance of the statistic is calculated for an random-censorship model which, however, does not require independence of the components of the censoring distribution. Two examples are given.