ABSTRACT
ABSTRACT In this chapter, some empirical Bayes monitoring techniques are presented for statistical process control (SPC). In particular, techniques are presented for monitoring univariate and multivariate continuous measurements, as well as yield/defect data (pass/fail data) and polytomous data. For each type of data, a Bayesian model is assumed for the process data, a prior with unknown hyperparameters is chosen, and empirical Bayes estimators are accordingly given for each of the parameters in the Bayesian model. In addition to estimating where the current process is, the estimators provide a distribution for the process parameter of interest, such as yield. Furthermore, by combining the empirical Bayes techniques with exponential smoothing, one can see where the process has been in the past, as well as where it is currently. Recursive equations for the estimators are provided for efficient computation, which is essential for successful implementation of real-time analysis of processes in factories.
