Data driven rank statistics in change point analysis

Abstract

The paper deals a class of rank based procedures for detection of changes with adaptively chosen scores. This is a certain continuation of the paper Antoch et al (2008). The limit behavior of the test procedures is studied both under the null as well as under a general class of alternatives. Accompanying simulation study focuses on various alternatives which are often met in practice. © 2009 Versita Warsaw and Springer-Verlag Berlin Heidelberg.