Tamakng University Department of Management Sciences
Two-variable gamma process with generalized Eyring model have been widely used to assess the reliability of reliable products in engineering applications. Because no close forms of the maximum likelihood estimators for the model parameters can be derived and iterative procedure to evaluate the maximum likelihood estimate is very sensitive to the initial input and difficult to control, the analytic genetic algorithm, Gibbs sampling Markov chain Monte Carlo algorithm and Metropolis-Hastings Markov chain Monte Carlo algorithm methods are established and applied to implementing parameter estimation for the gamma process. The performance of those methods are evaluated through simulations. Simulation results show that the Markov chain Monte Carlo method based maximum likelihood estimates outperform the other competitors with smaller bias and mean squared error. The application of the proposed methods was illustrated with a lumen degradation data set of light-emitting diodes.
International Journal of Information and Management Sciences 29(3), p.235-256