Article
Multidisciplinary Sciences
Wahyuni Suryaningtyas, Nur Iriawan, Heri Kuswanto, Ismaini Zain
Summary: The study introduces a new analytical model, the Hierarchical Bernoulli mixture model (Hibermimo), which combines the Bernoulli mixture with hierarchical structure data. By using the Hamiltonian Monte Carlo algorithm with a No-U-Turn Sampler, the model estimation shows that Hibermimo performs consistently well in modeling each district.
Article
Statistics & Probability
Emily Berg
Summary: The Iowa Seat-Belt Use Survey is an annual survey that provides estimates of seat-belt use rates in Iowa, United States. Small area estimation is necessary to obtain county-level estimates. Challenges arise from the multivariate count data and the observation of the same sampling units across five different time points. A unit-level Bayesian hierarchical model is used to address these challenges, incorporating multivariate dependencies and longitudinal data structure.
Article
Obstetrics & Gynecology
M. Ershadul Haque, Taslim Sazzad Mallick, Wasimul Bari
Summary: The study found that dropout from school is associated with lower attendance of ANC visits among women in Bangladesh. To ensure an adequate number of ANC visits for pregnant women, it is important to promote maternal education and prevent dropout after marriage.
BMC PREGNANCY AND CHILDBIRTH
(2022)
Article
Psychology, Multidisciplinary
Viet Hung Dao, David Gunawan, Minh-Ngoc Tran, Robert Kohn, Guy E. Hawkins, Scott D. Brown
Summary: Model comparison is crucial in psychological research, but existing methods have limitations in terms of computational cost and the number of models that can be compared. This paper proposes a novel algorithm, CVVB, which combines variational Bayes inference and Bayesian prediction for cross-validation, to address these issues. The results show that CVVB agrees strongly with model comparison using marginal likelihood but requires less time, enabling researchers to compare larger families of hierarchically specified cognitive models.
PSYCHOLOGICAL METHODS
(2022)
Article
Economics
Mikkel Bennedsen, Asger Lunde, Neil Shephard, Almut E. D. Veraart
Summary: This paper develops likelihood-based methods for estimation, inference, model selection, and forecasting of continuous-time integer-valued trawl processes. The full likelihood of integer-valued trawl processes is highly intractable, so a pairwise likelihood approach is used instead. The estimator obtained by maximizing the pairwise likelihood is shown to be consistent and asymptotically normal. The methods are also applied to financial bid-ask spread data, demonstrating the importance of carefully modeling the marginal distribution and autocorrelation structure.
JOURNAL OF ECONOMETRICS
(2023)
Article
Engineering, Mechanical
Junming Ma, Nani Bai, Yi Zhou, Chengming Lan, Hui Li, B. F. Spencer
Summary: This article proposes the use of generalized hierarchical Bayesian inference for fatigue life prediction based on general multi-parameter Weibull models. The article establishes a three-layer hierarchical Bayesian structure and uses Gibbs sampling to obtain posterior samples for parameters and hyperparameters. The results show that the scatter in fatigue life prediction for the corroded specimens becomes smaller when considering informative priors for the parameters in the Weibull model.
INTERNATIONAL JOURNAL OF FATIGUE
(2022)
Article
Physics, Multidisciplinary
Feng Zhao, Min Ye, Shao-Lun Huang
Summary: This paper studies the phase transition property of an Ising model on a stochastic block model, proposing a stochastic estimator for exact recovery and an unbiased convergent estimator for model parameters that can be computed in constant time. The stochastic algorithm can be transformed into an optimization problem, including special cases like maximum likelihood and maximum modularity. Metropolis sampling is used to verify the phase transition phenomenon through experiments.
Article
Computer Science, Artificial Intelligence
Hao Chen, Weikun Li, Weicheng Cui
Summary: Fitness functions of real-world optimization problems often require expensive experiments or numerical simulations for analysis. Integrating these into optimization algorithms directly leads to high computational costs. Surrogate-assisted evolutionary algorithms (SAEAs) have gained attention for their high efficiency in solving real world optimization problems. However, with the increase in dimension, the computational cost of constructing surrogates increases and their prediction accuracy may degrade. This paper proposes a surrogate-assisted evolutionary algorithm with hierarchical surrogate technique and adaptive infill strategy (SAEA-HAS) to address these challenges. Experimental results validate the effectiveness of SAEA-HAS.
EXPERT SYSTEMS WITH APPLICATIONS
(2023)
Article
Mathematics
Hua Xin, Jianping Zhu, Tzong-Ru Tsai, Chieh-Yi Hung
Summary: This study proposes a new three-statement randomized response estimation method to improve the estimation of sensitive-nature proportion (SNP) using hierarchical Bayesian modeling, Gibbs sampling, and Metropolis-Hastings algorithms. Monte Carlo simulations demonstrate that the method is effective and easy to use.
Article
Computer Science, Theory & Methods
Xiaohao Cai, Jason D. McEwen, Marcelo Pereyra
Summary: Bayesian model selection is a powerful method for objectively comparing models directly from observed data. However, it is computationally challenging due to the computation of the marginal likelihood. In this work, the authors propose a proximal nested sampling methodology that allows for the comparison of alternative Bayesian imaging models. The methodology is based on nested sampling and utilizes proximal Markov chain Monte Carlo techniques to efficiently handle large problems and specific model characteristics.
STATISTICS AND COMPUTING
(2022)
Article
Neurosciences
Yu Yao, Klaas E. Stephan
Summary: This article addresses the technical challenges of applying MCMC to hierarchical models for clustering in the space of latent parameters. Specifically focusing on dynamic causal models for fMRI and effective brain connectivity clustering, the study proposes a solution to improve convergence by introducing proposal distributions capturing dependencies between clustering and subject-wise generative model parameters. Validated on synthetic and real-world datasets, the proposed solution shows good convergence performance and superior runtime compared to state-of-the-art Monte Carlo techniques.
HUMAN BRAIN MAPPING
(2021)
Article
Social Sciences, Mathematical Methods
Robert G. Clark, David G. Steel
Summary: This paper presents a general approach for determining efficient sampling designs for probability-weighted maximum likelihood estimators in generalized linear models. The approach takes into account non-ignorable sampling, including outcome-dependent sampling. The new designs have probabilities of selection closely related to influence statistics such as dfbeta and Cook's distance. The effectiveness of the approach is demonstrated through a simulation based on data from the New Zealand Health Survey.
JOURNAL OF THE ROYAL STATISTICAL SOCIETY SERIES A-STATISTICS IN SOCIETY
(2022)
Article
Computer Science, Interdisciplinary Applications
Paul-Remo Wagner, Stefano Marelli, Bruno Sudret
Summary: In this paper, we propose a new Bayesian model inversion approach called SSLE, which approximates the likelihood function to obtain key statistics of the posterior distribution without being affected by the complexity of the forward model. The efficiency of SSLE is further enhanced by an adaptive sample enrichment scheme, making it suitable for a variety of challenging likelihood function problems.
JOURNAL OF COMPUTATIONAL PHYSICS
(2021)
Article
Ecology
Dave W. Hudson, Dave J. Hodgson, Michael A. Cant, Faye J. Thompson, Richard Delahay, Robbie A. McDonald, Trevelyan J. McKinley
Summary: This paper introduces a Bayesian approach to ecological system modeling, focusing on the task of multimodel inference through estimating posterior model weights. The Importance Sampling method used for model comparison aligns with RJ-MCMC model comparisons and is often more straightforward to fit and optimize.
METHODS IN ECOLOGY AND EVOLUTION
(2023)
Article
Statistics & Probability
David T. Frazier, Christopher Drovandi
Summary: Bayesian synthetic likelihood (BSL) is a well-established method for approximate Bayesian inference, but unreliable parameter inference can occur if the assumed data-generating process does not match the actual process. A new approach to BSL has been proposed to detect model misspecification and deliver accurate inferences even when the model is misspecified.
JOURNAL OF COMPUTATIONAL AND GRAPHICAL STATISTICS
(2021)