BAYESIAN-INFERENCE IN A DISCRETE SHOCK MODEL USING CONFOUNDED COMMON-CAUSE DATA

We consider redundant systems of identical components for which reliab ility is assessed statistically using only demand-based failures and s uccesses. Direct assessment of system reliability can lead to gross er rors in estimation if there exist external events in the working envir onment that cause two or more components in the system to fail in the same demand period which have not been included in the reliability mod el. We develop a simple Bayesian model for estimating component reliab ility and the corresponding probability of common cause failure in ope rating systems for which the data is confounded; that is, the common c ause failures cannot be distinguished from multiple independent compon ent failures in the narrative event descriptions.