Process monitoring in real time: Empirical bayes approach—discrete case

In this paper an empirical Bayes model is developed to monitor and analyse discrete data generated in a manufacturing process for printed circuit boards. A key feature of this analysis is the use of the current observation at time t and the posterior estimates of the distribution of the proportion nonconforming at time t – 1 to obtain a new, updated estimate of the posterior distribution at time t. The derived approach is widely applicable to statistical process control and provides a simple and fast algorithm for updating.