A0513
Title: Inference for multiple change-points in piecewise locally stationary time series
Authors: Wai Leong Ng - The Hang Seng University of Hong Kong (Hong Kong) [presenting]
Abstract: In a wide range of applications, such as econometrics, finance, and seismology, the stochastic properties of the observed time series change over time, and this phenomenon can be modelled by locally stationary time series models with time-varying parameter curves. However, the assumption of local stationarity may be violated. For example, abrupt changes in parameter values and the slope of the parameter curves may occur at some time points, referred to as jump points and kink points, respectively. In this case, piecewise, locally stationary time series models are more appropriate for modelling the stochastic properties of the time series. In contrast to the classical change-point setting in time series, methods for detecting both jumps and kinks in a piecewise locally stationary time series model are less developed. A three-step criterion-based procedure is presented for multiple change-point inferences in a piecewise locally stationary time series with possible jumps and kinks in its parameter curve. Theoretical properties are established, including consistency of the number and locations of the change-point estimation and the asymptotic exactness of the confidence intervals. Simulation studies and real data applications are provided to illustrate the performance of the proposed method.