Each week, we will be publishing layman’s abstracts of new articles from our prestigious portfolio of journals in statistics. The aim is to highlight the latest research to a broader audience in an accessible format.
The article featured today is from the Canadian Journal of Statistics, with the full article now available to read here.
Guo, L. and Modarres, R. (2020), Nonparametric change point detection for periodic time series. Can J Statistics, 48: 518-534. doi:10.1002/cjs.11545
This article considers the problem of detecting changes in a random sequence with periodic and autocorrelated distributions. Periodic changes blur the actual changes that are of interest. To adjust for the periodic effect, the article proposes to transform the sequence of periodic vector observations to matrix observations, which are free from the periodicity. Two detection statistics are proposed for different types of changes (changes in the mean and variance).
This paper describes an algorithm to detection a single change point in a periodic time series. It also proposes an algorithm to find multiple change points. Two hierarchical methods are used to specify the locations of the possible change points. A permutation test is given when that the total number of change points is not known. Simulations are conducted to compare different methods. The results show that the proposed methods outperform models that do not consider the periodicity. This paper can be used to detect changes in environmental, medical, and economic data, to name a few application areas.