Dissertation/Thesis Abstract

Robust Principal Component Analysis
by Kpamegan, Neil Racheed, M.S., American University, 2018, 47; 10784806
Abstract (Summary)

In multivariate analysis, principal component analysis is a widely popular method which is used in many different fields. Though it has been extensively shown to work well when data follows multivariate normality, classical PCA suffers when data is heavy-tailed. Using PCA with the assumption that the data follows a stable distribution, we will show through simulations that a new method is better. We show the modified PCA can be used for heavy-tailed data and that we can more accurately estimate the correct number of components compared to classical PCA and more accurately identify the subspace spanned by the important components.

Indexing (document details)
Advisor: Nolan, John P.
Commitee: Baron, Michael, Wall, Jane
School: American University
Department: Mathematics and Statistics
School Location: United States -- District of Columbia
Source: MAI 57/05M(E), Masters Abstracts International
Source Type: DISSERTATION
Subjects: Statistics
Keywords:
Publication Number: 10784806
ISBN: 9780355890495
Copyright © 2019 ProQuest LLC. All rights reserved. Terms and Conditions Privacy Policy Cookie Policy
ProQuest