2015 QPRC

CUSUM Algorithm for Detecting Translations in Gamma Distributions

Dan Jeske, University of California, Riverside

The CUSUM algorithm is developed for detecting translations of in-control gamma distributions. Efficient computing algorithms for the first two moments of the in-control and out-of-control run length distribution are provided. The special case of exponential distributions is discussed in some detail as it provides an unusual context where analytical formulae are available. The results are illustrated by applying them to a network surveillance context.