Simultaneous confidence regions for the parameters of damage processes (Q1324974)

In many cases the reliability of a product is affected by damage processes. These include processes of wear, corrosion and crack-growth. This paper describes statistical methods for confidence estimation of the parameters of the Wiener process as a model of the damage process. It is shown that for independent, not necessarily identically distributed observations of process increments, for observations of lifetime- distributions and for a mixture of these observations the assumptions of asymptotic normality of maximum-likelihood estimations (MLE) are fulfilled. The asymptotic normality of MLE is used to find simultaneous confidence regions for parameters of the damage process or, what is the same, for the resulting life-time distribution. Some ways to improve the convergence rate to asymptotic normality for small samples are developed.