Optimal detection of changing properties of random sequences. II: Sequential detection (Q1103576)

[For part I see the preceding review.] An optimal Bayesian sequential detection of a failure is considered under the assumption that the changes in the properties of the obseved process take place gradually and with a probability smaller than 1. In the case of independent observations the likelihood ratio averaged over the distribution of failure proves to be a sufficient statistic. The optimal stopping rules are found, and an illustrative example is given.