Article
Psychology, Mathematical
Don van den Bergh, Merlise A. Clyde, Akash R. Komarlu Narendra Gupta, Tim de Jong, Quentin F. Gronau, Maarten Marsman, Alexander Ly, Eric-Jan Wagenmakers
Summary: Linear regression analysis commonly involves two stages: defining the best model and using regression coefficients for prediction and evaluation; traditional inference methods often ignore model uncertainty, leading to overconfident parameter estimates. Model averaging is a technique that overcomes these drawbacks by weighting the contribution of each model for inference.
BEHAVIOR RESEARCH METHODS
(2021)
Article
Mathematical & Computational Biology
Ethan M. Alt, Matthew A. Psioda, Joseph G. Ibrahim
Summary: There is growing interest in using prior information in statistical analyses. In rare diseases, it can be challenging to establish treatment efficacy solely based on prospective study data due to small sample sizes. To address this issue, an informative prior can be elicited to determine treatment effects. This study develops a novel hierarchical prediction prior (HPP) that allows practitioners to elicit a prior prediction for the mean response in generalized linear models. The HPP approach shows higher efficiency gains compared to the conjugate prior and the power prior when predictions are incompatible with the data.
Article
Computer Science, Software Engineering
Jouni Helske
Summary: The R package walker extends Bayesian general linear models to handle time-varying effects of explanatory variables. This approach enables the modeling of intervention effects that gradually increase over time, such as changes in tax policy. The algorithm utilizes Hamiltonian Monte Carlo provided by Stan software to marginalize over the regression coefficients in a state space representation of the model, allowing for efficient low-dimensional sampling.
Article
Mathematical & Computational Biology
Jeffrey A. Boatman, David M. Vock, Joseph S. Koopmeiners
Summary: The increasing diversity of data sources offers more possibilities for estimating treatment effects, but borrowing must be done in a principled manner to reduce bias and errors. Estimators based on regression and Bayesian methods show promising performance in handling causal effects and can be applied to different data sources.
STATISTICS IN MEDICINE
(2021)
Article
Engineering, Electrical & Electronic
Prakash B. Gohain, Magnus Jansson
Summary: The Bayesian Information Criterion (BIC) is a commonly used criterion for model order estimation in linear regression models. However, it suffers from consistency and data scaling issues. This paper proposes a new form of BIC that addresses these problems and achieves consistency in both large sample sizes and high-SNR scenarios.
Article
Mathematics, Applied
Baria A. Helmy, Amal S. Hassan, Ahmed K. El-Kholy, Rashad A. R. Bantan, Mohammed Elgarhy
Summary: An entropy measure of uncertainty called extropy and its complementary function have gained attention in the past six years. However, it cannot be applied to long-lasting systems, leading to the concept of residual extropy. Bayesian and non-Bayesian estimators are used to estimate the extropy and residual extropy of the exponentiated gamma distribution. The performance of these estimators under different loss functions is evaluated.
Article
Health Care Sciences & Services
Federico Ricciardi, Silvia Liverani, Gianluca Baio
Summary: The regression discontinuity design is a quasi-experimental design that estimates the causal effect of a treatment when its assignment is defined by a threshold for a continuous variable. Bandwidth selection is an important decision in this design analysis, and the proposed methodology considers units' exchangeability as the main criteria for selecting subjects. The validity of the methodology is demonstrated through simulated experiments and an example on the effect of statins on cholesterol levels.
STATISTICAL METHODS IN MEDICAL RESEARCH
(2023)
Article
Automation & Control Systems
Jeremias Knoblauch, Jack Jewson, Theodoros Damoulas
Summary: The paper advocates for an optimization-centric view of Bayesian inference, introducing the Rule of Three (ROT) as a generalized method for Bayesian posteriors. It also explores the applications of Generalized Variational Inference (GVI) posteriors and their potential to improve robustness and posterior marginals in Bayesian Neural Networks and Deep Gaussian Processes.
JOURNAL OF MACHINE LEARNING RESEARCH
(2022)
Article
Multidisciplinary Sciences
Vera Lucia Damasceno Tomazella, Sandra Rego Jesus, Amanda Buosi Gazon, Francisco Louzada, Saralees Nadarajah, Diego Carvalho Nascimento, Francisco Aparecido Rodrigues, Pedro Luiz Ramos
Summary: This article introduces a method for estimating the parameters of the generalized normal linear regression model using Bayesian reference analysis, demonstrating its effectiveness through examples with artificial and real data on Eucalyptus clones from Brazil.
Article
Computer Science, Interdisciplinary Applications
Rahul Ghosal, Sujit K. Ghosh
Summary: Bayesian statistical inference for Generalized Linear Models (GLMs) with parameters lying on a constrained space can be challenging, especially when constructing valid prior distributions supported on a subspace spanned by a set of linear inequality constraints. A proposed generalized truncated multivariate normal prior distribution facilitates the construction of a general purpose product slice sampling method to obtain samples from the posterior distribution, showing computational efficiency for a wide class of GLMs with linear inequality constraints. The proposed method allows for easy quantification of uncertainty in parameter estimates and has shown superiority over existing methods in terms of sampling bias and variances in both simulated and real case studies.
COMPUTATIONAL STATISTICS & DATA ANALYSIS
(2022)
Article
Computer Science, Interdisciplinary Applications
Mei Li, Suthakaran Ratnasingam, Wei Ning
Summary: In this paper, three empirical likelihood (EL)-based methods are proposed to construct confidence intervals for quantile regression models with longitudinal data. Compared to traditional methods, the proposed methods exhibit better coverage performance, especially in small sample sizes.
JOURNAL OF STATISTICAL COMPUTATION AND SIMULATION
(2022)
Article
Automation & Control Systems
Jing Ouyang, Kean Ming Tan, Gongjun Xu
Summary: This paper proposes a debiasing approach for high-dimensional regression models with hidden confounding. By adjusting for the effects induced by the unmeasured confounders, the proposed method addresses the issue of invalidity in standard debiasing methods. The consistency and asymptotic normality of the proposed debiased estimator are established, and its finite sample performance is demonstrated through numerical studies and a genetic data set.
JOURNAL OF MACHINE LEARNING RESEARCH
(2023)
Article
Astronomy & Astrophysics
Marta Colleoni, Maite Mateu-Lucena, Hector Estelles, Cecilio Garcia-Quiros, David Keitel, Geraint Pratten, Antoni Ramos-Buades, Sascha Husa
Summary: In this study, the authors reanalyze the gravitational-wave event GW190412 using state-of-the-art phenomenological waveform models, focusing on the contribution from subdominant harmonics. They compare the PhenomX and PhenomT waveform models, discussing their construction techniques, computational efficiency, and agreement with other waveform models. Additionally, practical aspects of Bayesian inference, such as run convergence and computational cost, are also discussed.
Article
Mathematics
Xiaoning Li, Mulati Tuerde, Xijian Hu
Summary: This paper investigates the application of quantile regression models from a Bayesian perspective, proposing a hierarchical model framework and using Bayesian methods to handle missing data. The research findings demonstrate the significant advantages of the proposed methodology in both simulation and real data analysis.
Article
Computer Science, Theory & Methods
Zishu Zhan, Xiangjie Li, Jingxiao Zhang
Summary: Missing data is a common problem in clinical data collection, and this study focuses on the issue in the context of a semiparametric partially linear model. The choice of model structure is important, and a new imputation method called PRIME is proposed to handle incomplete data. When the correct model structure is unknown, PRIME-MA, which combines PRIME with model averaging, performs well. Simulations show that PRIME outperforms other methods, and a study on Chinese legal funding data confirms its effectiveness.
STATISTICS AND COMPUTING
(2023)
Article
Mathematics, Interdisciplinary Applications
Erengul Dodd, Jonathan J. Forster, Jakub Bijak, Peter W. F. Smith
Summary: The proposed method utilizes a negative binomial distribution and smoothness in parameter series to improve mortality projection accuracy, especially in areas of sparse data. By transitioning smoothly between GAMs and fully parametric models, it provides robust mortality projections across age groups.
SCANDINAVIAN ACTUARIAL JOURNAL
(2021)
Article
Social Sciences, Mathematical Methods
Petros Dellaportas, Evangelos Ioannidis, Christos Kotsogiannis
Summary: A modern system of Revenue Administration requires effective management of compliance, which involves a well-designed taxpayers audit strategy. The selection of taxpayers for audit is a non-standard sample size determination problem, involving initial random sampling and risk-based auditing to target taxpayers with the highest estimated risk in the population. The methodology in this paper aims to estimate the initial optimal random sample size for Revenue Administration Authorities to maximize expected tax revenues, illustrated using data from the UK's HMRC.
JOURNAL OF THE ROYAL STATISTICAL SOCIETY SERIES A-STATISTICS IN SOCIETY
(2021)
Article
Computer Science, Artificial Intelligence
Aristeidis Panos, Petros Dellaportas, Michalis K. Titsias
Summary: This study introduces a Gaussian process latent factor model for multi-label classification, which can capture correlations among class labels. To address computational challenges, several techniques are introduced, including variational sparse Gaussian process and stochastic optimization. The results demonstrate the practicality of this method in large-scale multi-label learning problems.
Article
Economics
Petros Dellaportas, Michalis K. Titsias, Katerina Petrova, Anastasios Plataniotis
Summary: In this paper, we introduce a multivariate stochastic volatility model that allows for flexible structure of the volatility matrix and treats all elements as functions of latent stochastic processes. We utilize a feasible and scalable MCMC method with quadratic computational complexity to perform inference. The model is applied to macroeconomic and financial data, demonstrating its effectiveness compared to other approaches.
JOURNAL OF ECONOMETRICS
(2023)
Article
Demography
Joanna Napierala, Jason Hilton, Jonathan J. Forster, Marcello Carammia, Jakub Bijak
Summary: This article suggests an alternative approach to asylum modeling by using statistical control theory to detect early warning signals, which can effectively assist in managing mixed migration flows, while further efforts are needed to fully utilize big data and scenario-based methods for comprehensive and systematic analysis of risk, uncertainty, and emerging trends.
INTERNATIONAL MIGRATION REVIEW
(2022)
Article
Hospitality, Leisure, Sport & Tourism
Evangelos Charamis, Christos Marmarinos, Ioannis Ntzoufras
Summary: The focus of this work is to estimate the number of team possessions in Euroleague basketball for multiple seasons and propose a new model for evaluating team performance. The study found a slight difference in the coefficient of free throw attempts that end a possession between Euroleague and NBA. By applying backward stepwise regression analysis, a new model, including true shooting percentage, offensive rebound percentage, and turnover percentage, was identified for evaluating European teams' performance.
INTERNATIONAL JOURNAL OF SPORTS SCIENCE & COACHING
(2023)
Article
Demography
Joanne Ellison, Ann Berrington, Erengul Dodd, Jonathan J. Forster
Summary: This study investigates the formal application of smooth methods implemented via generalized additive models (GAMs) to the analysis of fertility histories. It compares the performance of GAMs with existing methods and finds that smooth terms can offer considerable improvements in precision and efficiency.
DEMOGRAPHIC RESEARCH
(2022)
Article
Statistics & Probability
G. Tzoumerkas, D. Fouskakis, I Ntzoufras
Summary: The Power-Expected-Posterior (PEP) prior framework provides a convenient and objective method for variable selection in regression models from a Bayesian perspective. It inherits the advantages of the Expected-Posterior-Prior and avoids the need for selecting imaginary data. This study focuses on normal regression models with a smaller number of observations than explanatory variables, introduces the PEP prior methodology with different baseline shrinkage priors, and compares the results using simulated and real-life datasets.
JOURNAL OF STATISTICAL THEORY AND PRACTICE
(2022)
Article
Computer Science, Theory & Methods
Angelos Alexopoulos, Petros Dellaportas, Michalis K. Titsias
Summary: This paper introduces a general framework for constructing estimators with reduced variance for random walk Metropolis and Metropolis-adjusted Langevin algorithms. The proposed method uses the approximate solution of the Poisson equation to produce control variates, achieving variance reduction. Simulated data examples and real data examples verify the effectiveness of the method.
STATISTICS AND COMPUTING
(2023)
Article
Statistics & Probability
Jackie S. T. Wong, Jonathan J. Forster, Peter W. F. Smith
Summary: This study uses stochastic models to forecast mortality rates and introduces an improved APCI model with better fit and prediction accuracy compared to the Lee-Carter model. Model averaging is also performed to incorporate model uncertainty.
JOURNAL OF THE ROYAL STATISTICAL SOCIETY SERIES C-APPLIED STATISTICS
(2023)
Article
Statistics & Probability
Joanne Ellison, Ann Berrington, Erengul Dodd, Jonathan J. Forster
Summary: Fertility projections are important for anticipating demand for maternity and childcare services. Our Bayesian projection model combines individual-level data with aggregate population-level data, generating plausible forecasts that consider variables like educational qualifications, which may not be included in the population-level data.
JOURNAL OF THE ROYAL STATISTICAL SOCIETY SERIES C-APPLIED STATISTICS
(2023)
Proceedings Paper
Computer Science, Artificial Intelligence
Constantinos Daskalakis, Petros Dellaportas, Aristeidis Panos
Summary: This study provides guarantees for approximate Gaussian Process regression resulting from two common low-rank kernel approximations, and validates the effectiveness of the theoretical bounds through experiments.
THIRTY-SIXTH AAAI CONFERENCE ON ARTIFICIAL INTELLIGENCE / THIRTY-FOURTH CONFERENCE ON INNOVATIVE APPLICATIONS OF ARTIFICIAL INTELLIGENCE / THE TWELVETH SYMPOSIUM ON EDUCATIONAL ADVANCES IN ARTIFICIAL INTELLIGENCE
(2022)
Article
Mathematics, Interdisciplinary Applications
Angelos Alexopoulos, Petros Dellaportas, Omiros Papaspiliopoulos
Summary: This article takes a new look at solving the problem of volatility and jumps in daily stock returns and proposes a computational framework to improve the predictive ability of the models.
Article
Economics
Kai Hon Tang, Erengul Dodd, Jonathan J. Forster
Summary: The paper introduces a flexible and robust methodology for graduating mortality rates using adaptive P-splines, with an exponentially increasing penalty for sparse and unreliable data at high ages. By modeling male and female mortality rates jointly through penalties, information borrowing between the two sexes is achieved.
ANNALS OF ACTUARIAL SCIENCE
(2022)
Article
Social Sciences, Mathematical Methods
Jason Hilton, Erengul Dodd, Jonathan J. Forster, Peter W. F. Smith
Summary: The article explores the regularities of global mortality rates and models the changes in mortality rates in different countries by constructing a frontier mortality model, taking into account various factors that influence mortality. The findings help in predicting the trends in mortality rates over the next decade.
JOURNAL OF OFFICIAL STATISTICS
(2021)