Methods for assessing dependency between pairs of censored failure time variates will be described. These include estimators of parameters in semiparametric models, including Clayton's constant relative-risk model, and estimators of nonparametric dependency measures, including an average relative risk measure and a finite region version of Kendall's tau. Nonparametric estimators of the bivariate survivor function provide a basic tool for such assessment, as well as for a range of other multivariate failure time data analysis topics. A modification of the bivariate survivor function estimator of Dabrowska that removes negative mass and appears to improve estimation efficiency will be presented, along with preliminary work on a corresponding nonparametric maximum likelihood estimator. Regression generalizations of these various statistics will be briefly mentioned, along with genetic epidemiologic motivations.

Back to Workshop Schedule