CMStatistics 2021: Start Registration
View Submission - CMStatistics
B1537
Title: Profiled deviation subspaces with the application to change structure detection of high-dimensional data Authors:  Jiaqi Huang - Beijing Normal University (China) [presenting]
Wenbiao Zhao - Beijing Institute of Technology (China)
Xuehu Zhu - Xi'an Jiaotong university (China)
Lixing Zhu - Beijing Normal University (China)
Abstract: The focus is on dimension reduction of high-dimensional data onto the so-called profiled deviation subspaces such that we can equivalently work on lower-dimensional subspaces for detecting change structures within the sequence of data. Consistently estimating the dimensions of the subspaces is studied. Based on these studies, we propose a minimized criterion that is the minimum of component-wise criteria for the orthogonal components being in the subspaces. The estimation consistency of the number of changes and of their locations are verified with the investigation on what divergence rates of the original dimension of the data and the number of changes can ensure the above estimation consistencies. Numerical studies are conducted to examine the finite sample performance of the proposed method and two real data examples are analyzed for illustration.