![]() | Only 14 pages are availabe for public view |
Abstract Weighted exponential distribution has the probability density function whose shape is related closely to the shape of the probability density functions of Weibull, gamma and generalized exponential distribution. Therefore, we can use this model instead of any of these distributions. Weighted exponential distribution has been applied to a wide range of situations including applications in engineering, medicine and reliability. Accelerated life tests have widespread applications in manufacturing industries. Generally, information from tests at high levels of stress is extrapolated, through a physically reasonable statistical model, to obtain estimates of life or long term performance at normal levels of stress. The major assumption in accelerated life test is that the acceleration factor is known, or the life-stress relationships is known or can be assumed. In some cases, neither acceleration factor nor life-stress relationships are known and cannot even be assumed, that is, i.e., the data obtained from accelerated life test cannot be extrapolated to use conditions. Therefore, in such cases partially accelerated life test, which is a special case of accelerated life test, is a more suitable test to be performed, for which items are subject to both normal and accelerated conditions. The main purpose of this thesis is to make statistical inferences (estimation and prediction) for weighted exponential distribution constant-stress model under hybrid censoring scheme. The thesis contains three chapters presented as follows: In Chapter 1, some of essential definitions and concepts which will be used through-out this thesis are introduced. Also, Description of the model as well as previous studies is stated. At the end of this chapter we present description of the problem. In Chapter 2, estimation of the parameters of weighted exponential distribution is obtained based on constant-stress type-II hybrid censoring scheme. Maximum likelihood estimation is used for this purpose. Also, Bayes estimation is used under squared error loss and LINEX loss functions by using Markov chain Monte Carlo method. Approximate confidence intervals based on asymptotic normality of the maximum likelihood estimates of the parameters are constructed. Numerical simulation study is used to compare different estimates. In Chapter 3, the problem of predicting future items drawn from weighted exponential distribution with constant-stress partially accelerated life test, based on hybrid censoring scheme, is considered. We present one-sample and twosample prediction schemes. We construct predictive intervals for future observations of weighted exponential distribution items. Finally, Markov chain Monte Carlo method is used to find Bayesian predictive intervals. Simulation study has been executed and the results have been stated. |