서브메뉴
검색
Evaluating Forecasting Methods for Binary Outcomes with a Small Sample Size in Psychology
Evaluating Forecasting Methods for Binary Outcomes with a Small Sample Size in Psychology
상세정보
- 자료유형
- 학위논문 서양
- 최종처리일시
- 20250211151145
- ISBN
- 9798382629490
- DDC
- 151
- 저자명
- Luo, Lan.
- 서명/저자
- Evaluating Forecasting Methods for Binary Outcomes with a Small Sample Size in Psychology
- 발행사항
- [Sl] : The University of North Carolina at Chapel Hill, 2024
- 발행사항
- Ann Arbor : ProQuest Dissertations & Theses, 2024
- 형태사항
- 98 p
- 주기사항
- Source: Dissertations Abstracts International, Volume: 85-11, Section: B.
- 주기사항
- Advisor: Gates, Kathleen M.;Bollen, Kenneth A.
- 학위논문주기
- Thesis (Ph.D.)--The University of North Carolina at Chapel Hill, 2024.
- 초록/해제
- 요약Forecasting binary outcomes with small sample sizes presents unique challenges, particularly in the field of psychology where individual-level predictions are crucial. This dissertation evaluates various forecasting methods from traditional statistical models to machine learning approaches: ARMA, Elastic Net, Support Vector Machine, Random Forest, Extreme Gradient Boosting, Naive Bayes, and Long Short-Term Memory, in addition to two heterogeneous ensemble methods. Simulated data sets designed to mimic real-world psychological phenomena serve as the basis for testing each method's performance, focusing on their ability to forecast binary outcomes with a small number of observations from 20 to 60. Two empirical examples are further investigated to examine if the same inferences made based on the simulation study carry out and demonstrate practical applications. The findings suggest that while no single method universally outperforms others, certain models offer significant advantages depending on specific data characteristics and outcome variables. It underscores the importance of carefully considering the underlying data structure, the presence of class imbalance, the impact of nuisance covariates and missing data, and potential confounders, when choosing the most appropriate model.
- 일반주제명
- Quantitative psychology
- 일반주제명
- Computer science
- 일반주제명
- Clinical psychology
- 키워드
- Binary outcomes
- 기타저자
- The University of North Carolina at Chapel Hill Psychology
- 기본자료저록
- Dissertations Abstracts International. 85-11B.
- 전자적 위치 및 접속
- 로그인 후 원문을 볼 수 있습니다.
MARC
008250123s2024 us c eng d■001000017160983
■00520250211151145
■006m o d
■007cr#unu||||||||
■020 ▼a9798382629490
■035 ▼a(MiAaPQ)AAI31234916
■040 ▼aMiAaPQ▼cMiAaPQ
■0820 ▼a151
■1001 ▼aLuo, Lan.
■24510▼aEvaluating Forecasting Methods for Binary Outcomes with a Small Sample Size in Psychology
■260 ▼a[Sl]▼bThe University of North Carolina at Chapel Hill▼c2024
■260 1▼aAnn Arbor▼bProQuest Dissertations & Theses▼c2024
■300 ▼a98 p
■500 ▼aSource: Dissertations Abstracts International, Volume: 85-11, Section: B.
■500 ▼aAdvisor: Gates, Kathleen M.;Bollen, Kenneth A.
■5021 ▼aThesis (Ph.D.)--The University of North Carolina at Chapel Hill, 2024.
■520 ▼aForecasting binary outcomes with small sample sizes presents unique challenges, particularly in the field of psychology where individual-level predictions are crucial. This dissertation evaluates various forecasting methods from traditional statistical models to machine learning approaches: ARMA, Elastic Net, Support Vector Machine, Random Forest, Extreme Gradient Boosting, Naive Bayes, and Long Short-Term Memory, in addition to two heterogeneous ensemble methods. Simulated data sets designed to mimic real-world psychological phenomena serve as the basis for testing each method's performance, focusing on their ability to forecast binary outcomes with a small number of observations from 20 to 60. Two empirical examples are further investigated to examine if the same inferences made based on the simulation study carry out and demonstrate practical applications. The findings suggest that while no single method universally outperforms others, certain models offer significant advantages depending on specific data characteristics and outcome variables. It underscores the importance of carefully considering the underlying data structure, the presence of class imbalance, the impact of nuisance covariates and missing data, and potential confounders, when choosing the most appropriate model.
■590 ▼aSchool code: 0153.
■650 4▼aQuantitative psychology
■650 4▼aComputer science
■650 4▼aClinical psychology
■653 ▼aForecasting methods
■653 ▼aIndividual-level predictions
■653 ▼aHeterogeneous ensemble methods
■653 ▼aNuisance covariates
■653 ▼aBinary outcomes
■690 ▼a0632
■690 ▼a0984
■690 ▼a0622
■71020▼aThe University of North Carolina at Chapel Hill▼bPsychology.
■7730 ▼tDissertations Abstracts International▼g85-11B.
■790 ▼a0153
■791 ▼aPh.D.
■792 ▼a2024
■793 ▼aEnglish
■85640▼uhttp://www.riss.kr/pdu/ddodLink.do?id=T17160983▼nKERIS▼z이 자료의 원문은 한국교육학술정보원에서 제공합니다.
