Back to search

CSUniversity San Bernardino

Hellinger Distance-Based Methods for Change-Point Detection

Abstract

dc:description.abstract

Change-point analysis is used to identify locations in an ordered sequence where the underlying distribution changes. Such changes may occur in the mean, variance, shape, or other distributional features. Detecting these changes is important for understanding structural shifts in data and identifying regions where the statistical behavior of a process changes. In this thesis, we develop three change-point detection procedures based on the Hellinger distance: HD, \[M_{EXT}\], and \[M_{LOC}\]. The Hellinger distance is a well-established measure of discrepancy between probability distributions and is particularly useful because it is symmetric, bounded, and sensitive to general distributional differences. The proposed procedures are designed to detect changes without requiring a specific parametric form for the underlying distribution. Their finite-sample performance is examined through simulation studies under several distributional settings, including normal, skew-normal, exponential, Gamma, and \[t_5\] distributions. Empirical Type I error rates and statistical power are evaluated over different sample sizes, change-point locations, and types of distributional changes. The proposed procedures are compared with existing methods, including E-Divisive (ECP) and Pruned Exact Linear Time (PELT). The simulation results show that the proposed methods provide good Type I error control and competitive power across a wide range of settings. HD performs particularly well for changes in location, while \[M_{LOC}\] shows strong sensitivity to changes in variance and scale, and \[M_{EXT}\] provides competitive performance across several alternatives. Finally, the proposed methods are applied to breast cancer copy-number data to illustrate their ability to detect multiple structural changes in genomic sequences.

Degree

thesis:*
Name thesis:degree_name
Master of Arts in Mathematics
Level thesis:degree_level
Thesis
Discipline thesis:degree_discipline
Mathematics

Author and committee

dc:creator, dc:contributor.*
Author dc:creator
  • Van Sant, Angela M
Contributors dc:contributor
  • Suthakaran Ratnasingam

Subjects

dc:subject × 4

Identifiers

dc:identifier.*
Repository record dc:identifier
https://scholarworks.lib.csusb.edu/etd/2550

Chain of custody

source
Harvested from
CSUniversity San Bernardino
Base URL
scholarworks.lib.csusb.edu/do/oai/
Last updated
2026-11-20
Source record
OAI-PMH GetRecord
citation

Van Sant, Angela M. Hellinger Distance-Based Methods for Change-Point Detection. Thesis thesis, https://scholarworks.lib.csusb.edu/etd/2550