A0958
Title: Change-point estimation under epidemic alternatives
Authors: Soham Bonnerjee - University of Chicago (United States) [presenting]
Sayar Karmakar - University of Florida (United States)
George Michailidis - University of California, Los Angeles (United States)
Abstract: Traditional change-point analysis explores scenarios where the signal is piecewise. The purpose is to consider a situation where changes occur for a short time and come back to the original benchmark. Research on such epidemic patches or interval breakpoints is obscure in the existing literature and only discusses testing for the existence of such breaks. Detection accuracy results are established for the single and multiple epidemic scenarios. The naive algorithm has a lot of computational burden due to the nature of the problem. A novel block-based technique is proposed for detecting such breaks, and its theoretical guarantees are established. It is concluded by corroborating theoretical results with some simulation studies.
