Changepoint Detection Using Nonparametric Procedures

Abstract

This study investigates the detection of a changepoint due to location or scale or both location and scale changes in a sequence of univariate random variables. Two changepoint models, namely, the single change and the continuous change are considered and the appropriate test statistics are derived. The proposed test statistics are functions of two linear rank statistics, of which one is odd-translation invariant, sensitive to location shift and the other one is even-translation invariant, sensitive to scale shift. The asymptotic distributions of the proposed statistics are obtained under the null hypothesis.