A sequential multiple change-point detection procedure via VIF regression

摘要: In this paper, we propose a procedure for detecting multiple change-points in mean-shift model, where the number of is allowed to increase with sample size. A theoretic justification our new method also given. We first convert change-point problem into variable selection by partitioning data sequence several segments. Then, apply modified variance inflation factor regression algorithm each segment sequential order. When that suspected containing found, use weighted cumulative sum test if there indeed segment. The proposed implemented an which, compared two popular methods via simulation studies, demonstrates satisfactory performance terms accuracy, stability and computation time. Finally, analyze real examples.