서브메뉴
검색
Dropping the Anchor: Factor Analysis Models for the Simultaneous Assessment of DIF on Every Item as a Function of a Latent Covariate
Dropping the Anchor: Factor Analysis Models for the Simultaneous Assessment of DIF on Every Item as a Function of a Latent Covariate
상세정보
- 자료유형
- 학위논문 서양
- 최종처리일시
- 20250211150908
- ISBN
- 9798383690475
- DDC
- 151
- 서명/저자
- Dropping the Anchor: Factor Analysis Models for the Simultaneous Assessment of DIF on Every Item as a Function of a Latent Covariate
- 발행사항
- [Sl] : The University of North Carolina at Chapel Hill, 2024
- 발행사항
- Ann Arbor : ProQuest Dissertations & Theses, 2024
- 형태사항
- 107 p
- 주기사항
- Source: Dissertations Abstracts International, Volume: 86-02, Section: B.
- 주기사항
- Advisor: Bauer, Daniel J.
- 학위논문주기
- Thesis (Ph.D.)--The University of North Carolina at Chapel Hill, 2024.
- 초록/해제
- 요약Traditional methods to test for differential item functioning (DIF) within both the IRT and CFA frameworks have focused on between-group differences in the parameters linking the items to the latent trait of interest (e.g., intercepts and slopes). Less commonly, differences in item parameters are considered as a function of some observed continuous variable. Yet, there are potential sources of DIF that may not fit into these paradigms. One such context is when DIF emerges due to contamination from another trait that cannot be measured directly, but instead must be inferred from other observable variables. Further, traditional methods are limited to assessing DIF on subsets of items or single items in an iterative fashion, instead of assessing all items at the same time. A specific type of Confirmatory Factor Analysis (CFA) model, referred to as a nuisance CFA model, can be used to assess DIF without these limitations. However, it is vulnerable to a context-specific form of rotational indeterminacy and model under-identification called confound indeterminacy. With appropriate constraints this issue is remedied, as has been demonstrated in prior studies. In the current study, a novel form of the nuisance CFA model is presented that uses Bayesian estimation and informative priors to test potential DIF induced by a continuous latent variable, on every item simultaneously. The performance of this modeling approach is examined through a simulation study designed to compare the model's parameter recovery to other methods in both a traditional context where anchor items are used and contemporary contexts where different modeling constraints are used.
- 일반주제명
- Quantitative psychology
- 일반주제명
- Behavioral psychology
- 키워드
- Bias
- 키워드
- Psychometrics
- 기타저자
- The University of North Carolina at Chapel Hill Psychology
- 기본자료저록
- Dissertations Abstracts International. 86-02B.
- 전자적 위치 및 접속
- 로그인 후 원문을 볼 수 있습니다.
MARC
008250123s2024 us c eng d■001000017160107
■00520250211150908
■006m o d
■007cr#unu||||||||
■020 ▼a9798383690475
■035 ▼a(MiAaPQ)AAI30311964
■040 ▼aMiAaPQ▼cMiAaPQ
■0820 ▼a151
■1001 ▼aHamilton-Barlow, Atiyah J.
■24510▼aDropping the Anchor: Factor Analysis Models for the Simultaneous Assessment of DIF on Every Item as a Function of a Latent Covariate
■260 ▼a[Sl]▼bThe University of North Carolina at Chapel Hill▼c2024
■260 1▼aAnn Arbor▼bProQuest Dissertations & Theses▼c2024
■300 ▼a107 p
■500 ▼aSource: Dissertations Abstracts International, Volume: 86-02, Section: B.
■500 ▼aAdvisor: Bauer, Daniel J.
■5021 ▼aThesis (Ph.D.)--The University of North Carolina at Chapel Hill, 2024.
■520 ▼aTraditional methods to test for differential item functioning (DIF) within both the IRT and CFA frameworks have focused on between-group differences in the parameters linking the items to the latent trait of interest (e.g., intercepts and slopes). Less commonly, differences in item parameters are considered as a function of some observed continuous variable. Yet, there are potential sources of DIF that may not fit into these paradigms. One such context is when DIF emerges due to contamination from another trait that cannot be measured directly, but instead must be inferred from other observable variables. Further, traditional methods are limited to assessing DIF on subsets of items or single items in an iterative fashion, instead of assessing all items at the same time. A specific type of Confirmatory Factor Analysis (CFA) model, referred to as a nuisance CFA model, can be used to assess DIF without these limitations. However, it is vulnerable to a context-specific form of rotational indeterminacy and model under-identification called confound indeterminacy. With appropriate constraints this issue is remedied, as has been demonstrated in prior studies. In the current study, a novel form of the nuisance CFA model is presented that uses Bayesian estimation and informative priors to test potential DIF induced by a continuous latent variable, on every item simultaneously. The performance of this modeling approach is examined through a simulation study designed to compare the model's parameter recovery to other methods in both a traditional context where anchor items are used and contemporary contexts where different modeling constraints are used.
■590 ▼aSchool code: 0153.
■650 4▼aQuantitative psychology
■650 4▼aEducational tests & measurements
■650 4▼aEducational administration
■650 4▼aBehavioral psychology
■653 ▼aBayesian estimation
■653 ▼aBias
■653 ▼aConfirmatory Factor Analysis
■653 ▼aDifferential item functioning
■653 ▼aPsychometrics
■690 ▼a0632
■690 ▼a0288
■690 ▼a0514
■690 ▼a0384
■71020▼aThe University of North Carolina at Chapel Hill▼bPsychology.
■7730 ▼tDissertations Abstracts International▼g86-02B.
■790 ▼a0153
■791 ▼aPh.D.
■792 ▼a2024
■793 ▼aEnglish
■85640▼uhttp://www.riss.kr/pdu/ddodLink.do?id=T17160107▼nKERIS▼z이 자료의 원문은 한국교육학술정보원에서 제공합니다.
