Article
Psychology, Mathematical
Carl F. Falk, Todd A. Vogel, Sarah Hammami, Milica Miocevic
Summary: Mediation analysis helps understand how experimental manipulations change outcome variables. However, research on interval estimation for indirect effect in the 1-1-1 single mediator model is limited. Simulation studies on mediation analysis in multilevel data often do not match the typical scenarios encountered in experimental studies, and no study has compared resampling and Bayesian methods for constructing intervals in this context. Our simulation study compared the properties of interval estimates using bootstrap and Bayesian methods in the 1-1-1 mediation model. Bayesian credibility intervals performed well, but had lower power compared to resampling methods. The findings provide suggestions for selecting interval estimators for indirect effect based on the most important statistical property for a given study, and include R code for implementing all methods.
BEHAVIOR RESEARCH METHODS
(2023)
Article
Psychology, Mathematical
Xiao Liu, Zhiyong Zhang, Lijuan Wang
Summary: This study proposes a general approach to Bayesian hypothesis testing of mediation and demonstrates the impact of prior odds specifications on Bayesian hypothesis test of mediation.
BEHAVIOR RESEARCH METHODS
(2023)
Article
Mathematics, Interdisciplinary Applications
Camiel van Zundert, Emma Somer, Milica Miocevic
Summary: Bayesian mediation analysis requires specifying a prior for the covariance matrix of the variables involved. This paper introduces separation strategy priors and a Prior Predictive Check, providing guidelines for optimal prior specification based on researcher's prior knowledge encoding preferences.
STRUCTURAL EQUATION MODELING-A MULTIDISCIPLINARY JOURNAL
(2022)
Article
Mathematics, Interdisciplinary Applications
Tihomir Asparouhov, Bengt Muthen
Summary: This article discusses single and multilevel SEM models with latent variable interactions and demonstrates that the Bayesian estimation outperforms other methods, such as maximum-likelihood method, through simulation studies. Multilevel moderation models can be easily estimated with the Bayesian method.
STRUCTURAL EQUATION MODELING-A MULTIDISCIPLINARY JOURNAL
(2021)
Article
Mathematics, Interdisciplinary Applications
Sara van Erp, William J. Browne
Summary: Bayesian estimation of multilevel structural equation models has advantages in terms of sample size requirements and computational feasibility, but careful specification of the prior distribution, especially for random effects variance parameters, is necessary. The paper investigates alternative, more robust prior distributions for the doubly latent categorical multilevel model, highlighting the importance of constructing reasonable priors for multiple random effects variance parameters in MLSEMs. Although the robust priors outperform the traditional inverse-Gamma prior, consideration of hyperparameters is still crucial.
STRUCTURAL EQUATION MODELING-A MULTIDISCIPLINARY JOURNAL
(2021)
Article
Mathematics, Interdisciplinary Applications
Milica Miocevic, Shirin Golchi
Summary: Bayesian methods are often suggested for small sample research, but they require informative priors for better performance. This paper proposes an objective procedure for creating informative priors for mediation analysis based on historical data, leading to increased precision and power when data are exchangeable, without inducing bias when data are not exchangeable.
MULTIVARIATE BEHAVIORAL RESEARCH
(2022)
Article
Mathematics, Interdisciplinary Applications
Kelly D. Edwards, Timothy R. Konold
Summary: This study evaluates the performance of inaccurate (informative) priors in 1-1-1 multilevel structural equation modeling (MSEM) mediation under varying sample sizes, ICCs, and effect sizes. Results indicate that between-level indirect effect estimates are severely impacted, especially at small sample sizes.
STRUCTURAL EQUATION MODELING-A MULTIDISCIPLINARY JOURNAL
(2022)
Article
Mathematics, Interdisciplinary Applications
Milica Miocevic, Roy Levy, David P. MacKinnon
Summary: In the field of mediation analysis, Bayesian methods with informative priors can have different effects on statistical properties compared to frequentist methods, depending on sample size and the level of accuracy in the priors. It was found that the impact of inaccurate priors for loadings could be mitigated by reducing the informativeness of the prior, while this was not always the case for structural paths. The consequences of using informative priors varied depending on the inferential goals of the analysis, with inaccurate priors being more detrimental for accurately estimating the mediated effect.
MULTIVARIATE BEHAVIORAL RESEARCH
(2021)
Article
Psychology, Multidisciplinary
Christoph Koenig
Summary: The study aims to provide a new method for weighting informative prior distributions in Bayesian multiple regression models by combining sources of heterogeneity and a similarity measure ω. Through a comprehensive simulation study, the performance and behavior of the similarity-weighted informative prior distribution are investigated and compared to existing methods. The results offer applied researchers a means to specify accurate informative prior distributions.
FRONTIERS IN PSYCHOLOGY
(2021)
Article
Engineering, Mechanical
Teng Wang, Junchi Bin, Guillaume Renaud, Min Liao, Guoliang Lu, Zheng Liu
Summary: This paper proposes a new probabilistic prediction method that improves the prediction accuracy of fatigue crack growth by integrating multiple priors.
INTERNATIONAL JOURNAL OF FATIGUE
(2022)
Article
Psychology, Mathematical
Han Du, Brian Keller, Egamaria Alacam, Craig Enders
Summary: In this article, the DIC and WAIC criteria, widely used for Bayesian model assessment and comparison in Bayesian statistics, are studied. Different types of DIC and WAIC are compared using a multilevel mediation model as an example. The performance of conditional and marginal DICs and WAICs, as well as their performance with missing data, are investigated. The study finds that the marginal likelihood based DIC2, which excludes the likelihood of covariate models, generally has the highest true model selection rates.
BEHAVIOR RESEARCH METHODS
(2023)
Article
Psychology, Multidisciplinary
Steffen Zitzmann, Christoph Helm, Martin Hecht
Summary: Bayesian approaches are beneficial for estimating multilevel latent variable models in small samples, with prior distributions being used to overcome small sample problems by increasing the accuracy of estimation. This article discusses two approaches for specifying priors that aim at stabilizing estimators to reduce the MSE of the between-group slope estimator. Both approaches, involving slightly informative priors, have been shown to effectively reduce MSE in small samples, making them attractive options in such situations.
FRONTIERS IN PSYCHOLOGY
(2021)
Article
Computer Science, Information Systems
Ahmet Ulu, Gulcan Yildiz, Bekir Dizdaroglu
Summary: In this study, an attention-based CNN is introduced for image denoising. The network extracts texture information for detail preservation and utilizes multilevel feature extraction and weighting for noise removal. Experimental results demonstrate that the proposed method outperforms state-of-the-art denoising methods in terms of both quantitative and qualitative evaluations.
Article
Environmental Sciences
Jiming Zhang, Zheng Wang, Yiming Dai, Lei Zhang, Jianqiu Guo, Shenliang Lv, Xiaojuan Qi, Dasheng Lu, Weijiu Liang, Yang Cao, Chunhua Wu, Xiuli Chang, Zhijun Zhou
Summary: This study found an association between prenatal triclosan exposure and certain birth outcomes, possibly affecting intrauterine growth by interfering with thyroid-related hormones, gonadal hormones, adipokines, TMAO, and its precursors.
ENVIRONMENTAL RESEARCH
(2022)
Article
Mathematical & Computational Biology
Ethan M. Alt, Brady Nifong, Xinxin Chen, Matthew A. Psioda, Joseph G. Ibrahim
Summary: The study develops a method called "scale transformed power prior" for situations where historical and current data involve different data types. The scale transformed power prior addresses the issue of different data types by rescaling the parameter. Examples and simulation studies are provided to demonstrate the performance advantages of the scale transformed power prior over other priors.
STATISTICS IN MEDICINE
(2023)
Article
Mathematics, Interdisciplinary Applications
Yue Xiao, Hongyun Liu, Kit-Tai Hau
STRUCTURAL EQUATION MODELING-A MULTIDISCIPLINARY JOURNAL
(2019)
Article
Education & Educational Research
Dang Wang, Hongyun Liu, Kit-Tai Hau
Summary: The study successfully demonstrated the use of interactive simulation tasks in measuring critical thinking. Results showed reasonable reliability and validity, high student engagement with the game, but also highlighted students' hesitation in questioning others and applying critical thinking to problem-solving.
EDUCATION AND INFORMATION TECHNOLOGIES
(2022)
Article
Education & Educational Research
Kit-Tai Hau, Wen Jie Wu, Wing Tung Chung, Sze Ching Chan, Ming Ho Ng
Summary: With the COVID-19 outbreak, emergency remote teaching became the only alternative for schools. A large-scale survey conducted in Hong Kong showed concerns from teachers, principals, and parents about students' inability to concentrate and learn without teacher explanations. However, students, especially younger ones, perceived no worsening in academic achievement and felt more lively. Lack of computers and stable internet was not seen as a problem, and socially disadvantaged students did not differ in their perceived challenges or academic achievement.
EDUCATION AND INFORMATION TECHNOLOGIES
(2023)
Article
Health Policy & Services
Luyang Guo, Kit-Tai Hau
Summary: Using data from PISA 2018, this study examines the career expectations of adolescents in 61 economies regarding becoming doctors or nurses. The study found that 11% of adolescents aspired to be doctors, while only 2% aspired to be nurses. System-level factors like government health expenditure, safe work environment, and high salaries played a more significant role in attracting adolescents to healthcare careers compared to personal background factors.
HUMAN RESOURCES FOR HEALTH
(2023)
Article
Education & Educational Research
Leifeng Xiao, Kit-Tai Hau
Summary: We compared coefficient alpha with five alternatives in two simulation studies. Results showed that alpha performed well for unidimensional scales, but GLB and coefficient H overestimated reliability with small samples and short scales. For contaminated scales, all indices except omega h were reasonably unbiased with non-severe contamination, but alpha, omega total, and GLB were more sensitive in picking up contamination with shorter scales. Applied researchers should consider supplementary information of scale characteristics, avoid comparing different scales with one golden standard, and not use omega h alone.
APPLIED MEASUREMENT IN EDUCATION
(2023)
Article
Psychology, Multidisciplinary
Yuan Liu, Kit-Tai Hau, Xin Zheng
INTERNATIONAL JOURNAL OF PSYCHOLOGY
(2020)
Article
Education, Special
Che Kan Leong, Shek Tse, Wing Ki, Elizabeth Loh
INTERNATIONAL JOURNAL OF DISABILITY DEVELOPMENT AND EDUCATION
(2019)
Article
Psychology, Multidisciplinary
Yuan Liu, Kit-Tai Hau, Xin Zheng
JOURNAL OF PSYCHOLOGY
(2019)
Article
Education & Educational Research
Che Kan Leong, Mark Shiu Kee Shum, Chung Pui Tai, Wing Wah Ki, Dongbo Zhang
READING AND WRITING
(2019)
Article
Education & Educational Research
Dongbo Zhang, Keiko Koda, Che Kan Leong, Elizabeth Pang
JOURNAL OF RESEARCH IN READING
(2019)
Article
Education & Educational Research
Lihong Ma, Xiaofeng Du, Kit-Tai Hau, Jian Liu
EDUCATIONAL PSYCHOLOGY
(2018)
Article
Psychology, Multidisciplinary
Qishan Chen, Zhonglin Wen, Yurou Kong, Jun Niu, Kit-Tai Hau
FRONTIERS IN PSYCHOLOGY
(2017)
Article
Social Sciences, Interdisciplinary
Kevin Tze-wai Wong, Victor Zheng, Po-san Wan
SOCIAL INDICATORS RESEARCH
(2017)
Article
Education & Educational Research
Dongbo Zhang, Keiko Koda, Che Kan Leong
READING AND WRITING
(2016)