Abstract by Jackson Curtis
Modeling System Reliability using Bayesian Non-Parametric Methods
Data for system reliability is often in demand but is limited and expensive. There is a need for statistical methods to use component, subsystem, and other data to supplement the limited system data. A Bayesian framework is ideally suited for such a task. Additionally, utilizing the conjugate properties of some Bayesian nonparametric models provides computational advantages over parametric methods. In this work we explore the approximations made when combining components using these nonparametric methods. Fitting occurs using the first and second moment of the probability of failure. We find when the approximations are best, and when they breakdown.