서브메뉴
검색
Specifying Optimal Within-Subject Residual Variance-Covariance Structure in Latent Growth Model by Borrowing Power From Machine Learning
Specifying Optimal Within-Subject Residual Variance-Covariance Structure in Latent Growth Model by Borrowing Power From Machine Learning
상세정보
- 자료유형
- 학위논문 서양
- 최종처리일시
- 20250211153116
- ISBN
- 9798384462774
- DDC
- 379.1
- 저자명
- Yang, Junyeong.
- 서명/저자
- Specifying Optimal Within-Subject Residual Variance-Covariance Structure in Latent Growth Model by Borrowing Power From Machine Learning
- 발행사항
- [Sl] : The Ohio State University, 2024
- 발행사항
- Ann Arbor : ProQuest Dissertations & Theses, 2024
- 형태사항
- 130 p
- 주기사항
- Source: Dissertations Abstracts International, Volume: 86-04, Section: B.
- 주기사항
- Advisor: Kim, Minjung.
- 학위논문주기
- Thesis (Ph.D.)--The Ohio State University, 2024.
- 초록/해제
- 요약The latent growth model has been pivotal in understanding developmental processes for several decades. While most researchers have focused on growth factors' mean structure and variability, the within-subject residual variance-covariance structure has not received as much attention. The present study proposes a novel procedure for specifying a linear latent growth model's optimal within-subject residual variance-covariance structure. The method is based on the following ideas: approximating the true variance-covariance structure of the data, generating a series of replications with parameter values of the most probable within-subject residual variance-covariance structures within the data, and employing classification machine learning algorithms for prediction. A simulation examined the feasibility of the procedure predicting the true within-subject residual variance-covariance structure. Additionally, the performance of the proposed procedure was compared to the traditional approach of selecting models based on information criteria using the same replications. The results showed that the proposed method effectively detected true first-ordered autoregressive and banded main diagonal structures. In contrast, the information criteria, specifically the Bayesian Information Criterion and Sample-Size Adjusted BIC, effectively detected true identity and banded main diagonal structures, respectively. Based on these findings, suggestions were provided for researchers to consider when specifying the optimal within-subject structure of their data.
- 일반주제명
- Educational evaluation
- 일반주제명
- Educational psychology
- 일반주제명
- Quantitative psychology
- 일반주제명
- Computer science
- 키워드
- Feasibility
- 키워드
- Variability
- 기타저자
- The Ohio State University Educational Studies
- 기본자료저록
- Dissertations Abstracts International. 86-04B.
- 전자적 위치 및 접속
- 로그인 후 원문을 볼 수 있습니다.
MARC
008250123s2024 us c eng d■001000017165037
■00520250211153116
■006m o d
■007cr#unu||||||||
■020 ▼a9798384462774
■035 ▼a(MiAaPQ)AAI31693957
■040 ▼aMiAaPQ▼cMiAaPQ
■0820 ▼a379.1
■1001 ▼aYang, Junyeong.
■24510▼aSpecifying Optimal Within-Subject Residual Variance-Covariance Structure in Latent Growth Model by Borrowing Power From Machine Learning
■260 ▼a[Sl]▼bThe Ohio State University▼c2024
■260 1▼aAnn Arbor▼bProQuest Dissertations & Theses▼c2024
■300 ▼a130 p
■500 ▼aSource: Dissertations Abstracts International, Volume: 86-04, Section: B.
■500 ▼aAdvisor: Kim, Minjung.
■5021 ▼aThesis (Ph.D.)--The Ohio State University, 2024.
■520 ▼aThe latent growth model has been pivotal in understanding developmental processes for several decades. While most researchers have focused on growth factors' mean structure and variability, the within-subject residual variance-covariance structure has not received as much attention. The present study proposes a novel procedure for specifying a linear latent growth model's optimal within-subject residual variance-covariance structure. The method is based on the following ideas: approximating the true variance-covariance structure of the data, generating a series of replications with parameter values of the most probable within-subject residual variance-covariance structures within the data, and employing classification machine learning algorithms for prediction. A simulation examined the feasibility of the procedure predicting the true within-subject residual variance-covariance structure. Additionally, the performance of the proposed procedure was compared to the traditional approach of selecting models based on information criteria using the same replications. The results showed that the proposed method effectively detected true first-ordered autoregressive and banded main diagonal structures. In contrast, the information criteria, specifically the Bayesian Information Criterion and Sample-Size Adjusted BIC, effectively detected true identity and banded main diagonal structures, respectively. Based on these findings, suggestions were provided for researchers to consider when specifying the optimal within-subject structure of their data.
■590 ▼aSchool code: 0168.
■650 4▼aEducational evaluation
■650 4▼aEducational tests & measurements
■650 4▼aEducational psychology
■650 4▼aQuantitative psychology
■650 4▼aComputer science
■653 ▼aLatent growth model
■653 ▼aMachine learning algorithms
■653 ▼aFeasibility
■653 ▼aVariance-covariance structure
■653 ▼aVariability
■690 ▼a0632
■690 ▼a0288
■690 ▼a0443
■690 ▼a0525
■690 ▼a0984
■690 ▼a0800
■71020▼aThe Ohio State University▼bEducational Studies.
■7730 ▼tDissertations Abstracts International▼g86-04B.
■790 ▼a0168
■791 ▼aPh.D.
■792 ▼a2024
■793 ▼aEnglish
■85640▼uhttp://www.riss.kr/pdu/ddodLink.do?id=T17165037▼nKERIS▼z이 자료의 원문은 한국교육학술정보원에서 제공합니다.
