Statistics & Data Science GIDP Ph.D. Candidate
Image
When
Noon – 2 p.m., April 17, 2026
Where
Title: Epidemic Change Detection via Shifted Maximum Subarrays
Abstract: Change-point detection is a fundamental problem in modern statistics with broad applications in genomics, epidemiology, and signal processing. An important variant is the epidemic change-point problem, which concerns detecting transient changes that appear and disappear within a sequence. Despite its relevant in modeling local structural shifts, such as temporary outbreaks or genomic aberrations, accurately identifying epidemic changes and assigning valid p-values remain challenging both theoretically and computationally. In this report, we focus on efficient computation of the scan statistic raised in epidemic change-point problem. Specifically, we introduce the Shifted Maximum Subarray (SMS) problem and characterize its solution path. More importantly, we prove that the scan statistic can be derived directly from the SMS solution path. Consequently, we develop an empirically near-linear-time algorithm to compute the exact value of the scan statistic. In addition, we incorporate our algorithm into a nonparametric permutation testing framework and demonstrate its effectiveness in both Type I error control and statistical power.