الفهرس 
Bayesian Prediction Intervals Based on ProgressivelyOnly 14 pages are availabe for public view
 |  | from | 185 |  |  |
| | | from | 185 | | |
Abstract
Prediction is one of the most important issues in statistical inference and it is equally important and useful as statistical estimation. Meteorology, medicine, economics, finance, engineering, politics and education are applied disciplines in which rediction is essential and is therefore of greet interest. In many practical problems, one would wish to use the information from the past sample to predict the observations of a future sample from the same population with a specified probability. One way to do this is to construct an interval that will contain the future values of interest, and such an interval is called a prediction interval. Prediction intervals are of different types which include onesample prediction, two-sample prediction, and multi-sample prediction. It is common in life-testing and reliability experiments that some experimental units are either lost or removed from experimentation before their failure. The loss may occur unintentionally, or it may have been designed in the study. In this case, we will obtain only a portion of the sample information. Data obtained from such experiments are called censored data. Censored data are commonly ncountered in reliability theory, survival analysis and clinical trials. The main subject of this thesis is the Bayesian prediction interval problem. Here, we discuss this problem based on different forms of censored data, namely, (1) Type-II right censored data, (2) multiply Type-II censored data, (3) progressively Type-II censored data, (4)
Type-I hybrid censored data, and (5) Type-II hybrid censored data. In each case, we use a general form for the underlying distribution and a general conjugate prior to develop a very general procedure for determining one- and two-sample Bayesian prediction intervals for future life-lengths. We then present the results for some specific continuous distributions as illustrative examples. Finally, we carry out a simulation study to evaluate the performance of all the prediction procedures
developed here.