Cost-optimal multistage sampling plans in statistical quality control

Multistage Bayesian decision procedures in statistical quality control are known from attribute sampling. In this paper they are introduced in a more general framework occuring in lot-control by using the theory of Bayesian sequentially planned decision procedures. We show that under sufficiency and transitivity assumptions and monotonicity properties concerning the distribution and cost set-up these Bayes-procedures have (z, c ^–,c ⁺)-structure which, on one hand, generalizes results of K.-H. Waldmann and, on the other hand, reduces computational effort significantly. Finally, examples taken from attribute sampling and life testing for an outgoing lot are presented.The theory presented in this paper is taken from 11].