ABSTRACT
This chapter is devoted to sequential non-Bayesian changepoint detection procedures and study of their optimality and asymptotic optimality properties. We first discuss analytic formulas for investigating the properties of a number of on-line detectors, namely Shewhart control charts, geometric moving average charts, finite moving average charts, CUSUM-type algorithms, and the GLR detector. We then discuss optimality properties of the Shiryaev-Roberts-type detection procedures. We also discuss numerical methods for solving integral equations which are useful for estimating operating characteristics of change detection algorithms. Second, we present the results of comparison between different algorithms using analytical methods, numerical techniques and statistical simulations. We also discuss robustness issues of some algorithms with respect to a priori information.
