View : 476 Download: 0

Full metadata record

DC Field Value Language
dc.contributor.advisor유재근-
dc.contributor.author유혜민-
dc.creator유혜민-
dc.date.accessioned2025-02-18T16:32:25Z-
dc.date.available2025-02-18T16:32:25Z-
dc.date.issued2025-
dc.identifier.otherOAK-000000243172-
dc.identifier.urihttps://dcollection.ewha.ac.kr/common/orgView/000000243172en_US
dc.identifier.urihttps://dspace.ewha.ac.kr/handle/2015.oak/271015-
dc.description.abstract충분 차원 축소(sufficient dimension reduction)의 영역에서, 전통적인 모멘트 기반 방법인 SIR(Sliced Inverse Regression)과 SAVE(Sliced Average Variance Estimates)는 널리 사용되었다. 그러나 이러한 방법들은 변수 간의 복잡한 관계를 포착하는 데 있어 유연성이 부족한 등의 한계를 가지고 있다. 이러한 한계를 해결하기 위해 다양한 대안적 접근법들이 개발되었다. 본 논문은 이러한 대안적 방법들 중 두 가지, 즉 우도 기반 충분 차원 축소 기법인 LAD(Likelihood Acquired Directions)와 UPFC(Unstructured Principal Fitted Components)의 성능을 비교하는 데 중점을 둔다. 두 방법은 공통적으로 우도 원리를 기반으로 하지만, 기본 가정, 데이터 구조 모델링 방식 및 계산 절차 등에서 차이를 보인다.;In the realm of sufficient dimension reduction, traditional moment-based methods such as sliced inverse regression (Li 1991) and sliced average variance estimation (Cook and Weisberg 1991) have been widely utilized. However, these methods come with inherent limitations, such as reduced flexibility in capturing complex relationships between variables. To address these shortcomings, various alternative approaches have been developed. This paper focuses on comparing two advanced methods—likelihood acquired directions (Cook and Forzani 2009) and unstructured principal fitted components (Cook and Forzani 2008)—both of which are likelihood-based sufficient dimension reduction techniques. While these methods share a common foundation in likelihood principles, they differ in their underlying assumptions, how they model data structures, and computational procedures.-
dc.description.tableofcontentsI. Introduction 1 II. Likelihood-Based Dimension Reduction 3 A. Likelihood Acquired Directions 4 B. Unstructured Principal Fitted Components 5 III. Comparison of LAD and UPFC 8 A. Theoretical Foundation 8 B. Model Assumptions 9 C. Handling of Response Variable 10 D. Estimation Procedures. 11 E. Interpretation of Log-Likelihood Terms 12 F. Robustness 13 IV. Simulations 15 A. Constant Normal Error Variance 16 B. Non-Normal Error Variance 17 C. Non-Constant Error Variance 18 V. Discussion 20 Bibliography 21 Abstract(in Korean) 23-
dc.formatapplication/pdf-
dc.format.extent432518 bytes-
dc.languageeng-
dc.publisher이화여자대학교 대학원-
dc.subjectSufficient dimension reduction, Likelihood acquired directions, Unstructured principal fitted components-
dc.subject.ddc500-
dc.titleComparative Analysis of Likelihood-Based Dimension Reduction Methods-
dc.typeMaster's Thesis-
dc.creator.othernameYoo, Hyemin-
dc.format.pageiii, 23 p.-
dc.identifier.thesisdegreeMaster-
dc.identifier.major대학원 통계학과-
dc.date.awarded2025. 2-
Appears in Collections:
일반대학원 > 통계학과 > Theses_Master
Files in This Item:
There are no files associated with this item.
Export
RIS (EndNote)
XLS (Excel)
XML


qrcode

BROWSE