Abstract: The cumulative sum (CUSUM) chart, well-known to be sensitive in detecting small and moderate parameter changes, is proposed here for monitoring a high yield process. The sensitivities of the CUSUM charts based on geometric, Bernoulli and binomial counts are compared. Based on the comparisons, recommendations for the selection of a chart are provided. Simple procedures are given for optimal design of CUSUM charts based on geometric and Bernoulli counts. An application of CUSUM charts in monitoring an actual high yield process is demonstrated.
Key words and phrases: Bernoulli counts, geometric counts, Markov chain, parts-per-million, statistical process control.