Statistics and Its Interface

Volume 13 (2020)

Number 2

A variable selection approach to multiple change-points detection with ordinal data

Pages: 251 – 260

DOI: https://dx.doi.org/10.4310/SII.2020.v13.n2.a9

Authors

Chi Kin Lam (Department of Statistics and Actuarial Science, University of Hong Kong)

Huaqing Jin (Department of Statistics and Actuarial Science, University of Hong Kong)

Fei Jiang (Department of Epidemiology and Biostatistics, University of California, San Francisco, Cal., U.S.A.)

Guosheng Yin (Department of Statistics and Actuarial Science, University of Hong Kong)

Abstract

Change-point detection has been studied extensively with continuous data, while much less research has been carried out for categorical data. Focusing on ordinal data, we reframe the change-point detection problem in a Bayesian variable selection context. We propose a latent probit model in conjunction with reversible jump Markov chain Monte Carlo to estimate both the number and locations of change-points with ordinal data. We conduct extensive simulation studies to assess the performance of our method. As an illustration, we apply the new method to detect changes in the ordinal data from the north Atlantic tropical cyclone record, which has an indication of global warming in the past decades.

Keywords

latent variable, multiple change-points, ordinal data, probit model, reversible jump Markov chain Monte Carlo

The authors’ research was supported in part by a grant (no. 17307218) from the Research Grants Council of Hong Kong.