A General Framework For Constructing Locally Self-Normalized Multiple-Change-Point Tests

TL;DR

This paper introduces a flexible, self-normalized framework for detecting multiple change points in time series data, eliminating the need for nuisance parameter estimation and outperforming existing methods in accuracy and robustness.

Contribution

It presents a novel, general approach that unifies various change-point detection methods without requiring parameter tuning or pre-specified number of change points.

Findings

Framework is size-accurate and robust in finite samples

Outperforms existing methods in power and robustness

Validated with case studies on financial data

Abstract

We propose a general framework to construct self-normalized multiple-change-point tests with time series data. The only building block is a user-specified one-change-point detecting statistic, which covers a wide class of popular methods, including cumulative sum process, outlier-robust rank statistics and order statistics. Neither robust and consistent estimation of nuisance parameters, selection of bandwidth parameters, nor pre-specification of the number of change points is required. The finite-sample performance shows that our proposal is size-accurate, robust against misspecification of the alternative hypothesis, and more powerful than existing methods. Case studies of NASDAQ option volume and Shanghai-Hong Kong Stock Connect turnover are provided.