|
Abstract
|
In the presence of independent competing risks, high-reliability and long-lifetime units are usually tested using accelerated life testing method. The estimates of reliability characteristics are obtained based on a known acceleration model and a latent competing risks model. For the cases with unknown acceleration models and dependent competing risks, we introduce a copula based partially accelerated competing risks model with a tampered random variable transformation between the use lifetime and the accelerated lifetime of each failure cause. For this model, we derive the reliability function and dependence structure of the partially accelerated competing risks data. In consideration of the parametric constraints and complicated joint posterior distribution, we define unconstrained parameters by a specific transformation function, and utilize the Hamiltonian Monte Carlo method within MCMC procedures for the estimation of model parameters. A simulation study is conducted to investigate the estimation performance, and a real data analysis is presented for further illustration of reliability estimates. The performance indicates the feasibility of the constrained parametric Bayesian estimation method for the proposed model.
|