Bayes Estimation For Accelerated Life Tests Under Type ? Progressive Censoring With Random Removals

Assume that the life of product follows Gumbel distribution,under the Type I model,constant stress is used in the test.The test process using continuous observation,and there are no failure products are removed at random.,the numbers of it obey the binomial distribution.In this paper,the maximum entropy principle is used to determine the prior distribution.When the stress level number is small,sample allocation proportion is selected as a optimization variable,and square loss function as optimization criterion,in the minimum square loss function to achieve sample optimal allocation proportion.To stress level number is large,using M-H algorithm to estimate the model parameters,and the mean and standard deviationget of parameters for bayesian updates,the final parameter estimates will gain after update multiple,and compare it with the maximum likelihood estimate.Finally,using simulated random life data for instance analysis.