Nonparametric methods in change-point problems

From MaRDI portal





This monograph is on a subject of rapidly growing interest. In its simplest form the problem considered is as follows: suppose that \(\xi_ 1,\dots,\xi_ n\) is a sequence of independent random variables such that for some \(0<\theta<1\), \(\xi_ 1,\dots,\xi_{[n\theta]}\) have distribution \(F\) and \(\xi_{[n\theta]+1},\dots,\xi_ n\) have distribution \(G\), with \(F\neq G\). The goal of statistical inference then is to estimate the unknown change-point \(\theta\). Several extensions of this `abrupt change' model are discussed: gradual, disappearing and multiple disorders. In each case it is assumed that disorders occur due to changes in the mean. Accordingly, the detection procedures are based on partial sum processes associated with the data. Strong consistency and distributional convergence of the estimators are shown. In most situations the assumption of independence is replaced by a weaker mixing condition. Apart from a review chapter, the monograph contains four chapters entitled: a posteriori change-point problems, sequential change-point detection problems, disorder detection of random fields, applications of nonparametric change-point detection methods. To this reader the book would have benefitted from postponing the study of weakly dependent data to latter sections and first elaborating the ideas associated with change-point detection for independent data. Thus, from a pedagogical viewpoint, the material is not arranged in an optimal manner. Sometimes the English translation is not smooth. Though, if a reader is willing to reread Neumann-Pierson as Neyman-Pearson (p. 25) and does not worry about statements like, e.g., `a Brownian Bridge has independent increments' (p. 3), he will find the book quite useful. The bibliography is extensive. No excercises and no index.




Cited in
(68)








This page was built for publication: Nonparametric methods in change-point problems

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q3136820)