A study of shifting models in life tests via Bayesian approach using semi-or-used priors (SOUPS) (Q5902940)

Consider a life test where the observations are obtained from the general exponential model. It is assumed that a parameter of the model shifts at some stage of the sampling procedure; both single and multiple shifts are considered. A Bayesian approach is adopted and the effect of these shifts in parameter on the predictive distribution of future observations is discussed. The initial prior is updated to a new prior at the shift point by using the immediately preceding estimate of the parameter (termed a semi-or-used prior) instead of assigning a new prior at this point.