Abstract: The cumulative incidence function provides intuitive summary information about competing risks data. Via a mixture decomposition of this function, we study how covariates affect the cumulative incidence probability of a particular failure type at a chosen time point. Without specifying the corresponding failure time distribution, several inference methods are constructed based on imputation and weighting approaches. Large sample properties of the proposed estimators are derived, and their finite sample performances are examined via simulations. For illustrative purposes, the proposed methods are applied to well-known heart transplant data and compared with the analysis of Larson and Dinse (1985). In the on-line Supplement, we also apply our methods to analyze the Taiwan nationwide laboratory-confirmed severe acute respiratory syndrome (SARS) database.
Key words and phrases: Cause-specific hazard, cumulative incidence function, imputation, inverse probability of censoring, logistic regression, missing data, mixture model.