الفهرس 
Bayesian prediction of future observations from finite mixture distributions under random censoring
AbstractOnly 14 pages are availabe for public view
 |  | from | 123 |  |  |
| | | from | 123 | | |
Abstract
Bayesian analysis has been extensively used for many applications in reliability. In various situations mixture survival models is a necessity for modeling a wide range of observed phenomena which do not normally yield to modeling through classical distributions. This study focuses on the Bayesian prediction of future observations coming from a population modeled by a two-component mixture survival model. Specifically, a random sample is drawn from the population and is subjected to random censoring which is a generalization of type I censoring. Based on the data available from this sample, the thesis aims at obtaining interval and point Baysian prediction of future observation and kth ordered observation from a second independent sample (two-sample prediction). Observations under study are assumed to be modeled by a two-component general survival model. The general model includes as special cases several survival models which are extensively applied in survival analysis. The study includes two cases. The first is the case where that two-component general model has three unknown parameters, one for each component and the third is the mixing proportion. This will be applied to two component Weibull (two-component exponential and two-component rayleigh as special cases) and two component burr type XII (two-component momax as a special case). A numerical example is presented for each of these special cases. A simulation study is done when the model is a two-component exponential