Model Mis‐Specification Analysis for Random Effect and Covariate Based on Wiener Degradation Process

ABSTRACT Thanks to the extensive applications of accelerated degradation tests (ADT), the influence of covariates has been paid more and more attention. We propose a power‐law form, which is studied with other previously perceived covariate types for model uncertainty not neglecting unidentified distribution of drift parameter. It simultaneously determines the distribution of volatility parameters under Wiener process (WP) when the dependency inner model's parametric space exists. The Markov Chain Monte Carlo (MCMC) algorithm implemented by WinBUGS is employed to deal with general failure possibility functions of multiple hypothesis combinations. A new criterion aimed at measuring the volatility variable's goodness‐of‐fit is involved a company with several Akaike information criterion (AIC)‐like indexes for an overall model comparison and selection. It is convincing that after several numerical and real example analyses, the interaction between the covariate and latent variable distribution influences the performance of computation to a large degree.