A Bivariate Exponential Model with Covariates in Competing Risk Data

Abstract

Sarkar's bivariate exponential distribution is incorporated in a competing risk model with two causes of failure. In view of the nonidentifiability of the parameters of this distribution under competing risk, reparameterization is advocated and covariates are related to the reparameterized parameters through logistic and log­linear models. The restricted alternative hypothe,is is considered for the comparison of the survival distributions of the two causes of failure, the test statistic based on Roy's union­intersection principle is used and compared with the score test statistic. An application is also considered,