An Efficient Algorithm For Maximum Likelihood Estimation In Semiparametric Degradation Models

A number of flexible semiparametric models have been proposed for degradation data in recent years.Because of excessive numbers of model parameters,parameter estimation in these models turns out to be difficult tasks.The EM algorithm is a common choice for maximum likelihood estimation in these models.When the inspection times of the sample units distinct from each other,however,the convergence rate of the EM algorithm has been found to be extremely slow.To overcome the difficulty,a new algorithmic framework,which is based on a modification of the Generalized Rosen(GR)algorithm,is proposed for the semiparametric estimation.Convergence properties of the algorithm are investigated,and detailed implementations of the algorithm for the stochastic degradation processes are developed.A comprehensive simulation study shows that the proposed algorithm significantly outperforms the EM algorithm in terms of both accuracy and efficiency.A liner wear data set is used to illustrate the developed algorithm.