Title: Weighted adaptive hard threshold signal approximation for robust change point detection
Authors: Xiaoli Gao - University of North Carolina at Greensboro (United States) [presenting]
Abstract: The copy number detection is incorporated into a change point regression model and a new robust change point detection method is proposed to simultaneously identify breakpoints and outliers. This new model incorporates an individual weight for each observation and uses the adaptive hard threshold approach to efficiently locate both outliers and CNVs. A novel way to select tuning parameters is also adopted in this model. The performance of the proposed change point method is demonstrated by both simulations and real data analysis. We show that the new model preforms more accurately in detecting the number of true break points, particularly given noisy data sets. As a by-product, the proposed approach can detect outliers efficiently.