Competing Risk Data Generation Based on Subdistribution

Abstract

Competing risks data arises when individual may fail due to one of several possible causes called competing risks. The algorithm for generating such data is needed for simulation purposes. We propose a method to generate such data based on the subdistribution function. This method can be considered general to overcome different kind of subdistribution function. The algorithm is developed in three steps: (1) generating the type of failure, (2) generating survival time following conditional subdistribution, and (3) generating censoring time. The percentage of censoring is determined analytically. Example and implementation of the algorithm in statistical package R is provided. The generated data succesfully satisfy the intended properties.