574 – Optimal Portfolios, Clustering, and Change Points
A New Algorithm for Multiple Change-Points Estimation in Time Series
Ngai Hang Chan
Chinese University of Hong Kong
Chun-Yip Yau
Chinese University of Hong Kong
Rongmao Zhang
Zhejiang University
We consider the structural break autoregressive process where a time series has an unknown number of break-points, and the time series follows a stationary AR model in between any two break-points. It is well-known that the estimation of the locations of the break-points involves huge computational challenges. By reformulating the problem in a regression variable selection context, we propose in this paper a group least absolute shrinkage and selection operator (LASSO) procedure to estimate the number and the locations of the break-points, where the computation can be efficiently performed. Simululation studies are conducted to assess the finite sample performance.