A0216
Title: A Bayesian multivariate Student-t degradation model for dependent log-increments
Authors: I-Tang Yu - Tunghai University (Taiwan) [presenting]
Raf Loreto - Tunghai University (Italy)
Abstract: The aim is to propose the use of the multivariate Student-t distribution to describe log-increments of degradation measures within an experimental unit. This approach captures both the dependence between the log increments and the heavy-tailed behavior, which can be observed in real-world data. Within the Bayesian analysis framework, an MCMC algorithm with the use of a Gibbs sampler is developed to estimate the model parameters. Lifetime predictions are then made using the Monte Carlo method due to the lack of the additive property of the corresponding log-Student-t distribution. To demonstrate the practical application of the model, two examples are analyzed. The results highlight the robustness of the proposed model to assumption variations, which is confirmed through a simulation study.
