Article
Economics
Fabio Canova, Christian Matthes
Summary: In this article, we discuss the use of composite likelihood function in dynamic stochastic general equilibrium models to address issues related to estimation, computation, and inference. By combining information from different models or datasets, we are able to estimate common parameters and provide alternative interpretations for the methodology used. Various examples are presented to demonstrate the potential of this approach in resolving well-known problems and justifying the pooling of different estimates.
Article
Statistics & Probability
Haben Michael, Yifan Cui, Scott A. Lorch, Eric Tchetgen J. Tchetgen
Summary: This article introduces Marginal Structural Models (MSMs), a type of counterfactual models for studying the joint effects of time-varying treatment regimes in complex longitudinal studies subject to time-varying confounding. The author establishes identification of MSM parameters under a Sequential Randomization Assumption (SRA), which assumes no unmeasured confounding of treatment assignment over time. When the Sequential Randomization Assumption fails, we propose using a time-varying instrumental variable to identify the parameters of a subset called Marginal Structural Mean Models (MSMMs). The article presents a weighted estimator and evaluates its performance through simulation studies, applying it to investigate the effect of delivery hospital type on neonatal survival probability.
JOURNAL OF THE AMERICAN STATISTICAL ASSOCIATION
(2023)
Article
Multidisciplinary Sciences
Taehwa Choi, Hyunjun Lee, Sangbum Choi
Summary: Dynamic treatment regime (DTR) is an emerging paradigm in recent medical studies that assigns optimal treatments to patients based on individual features such as genetics, environment, and social factors. This article proposes an algorithm for estimating optimal dynamic treatment regimes with survival endpoints using a doubly-robust weighted classification scheme. The algorithm simplifies the estimation procedure by utilizing a pseudo-value approach for censored survival data, allowing standard machine learning techniques to be applied without losing efficiency. The algorithm is further enhanced with SCAD-penalization and modified algorithms to handle multiple treatment options.
SCIENTIFIC REPORTS
(2023)
Article
Biochemical Research Methods
Sebastian Persson, Niek Welkenhuysen, Sviatlana M. Shashkova, Samuel R. Wiqvist, Patrick M. Reith, Gregor R. Schmidt, Umberto M. Picchini, Marija R. Cvijovic
Summary: Understanding the causes and consequences of heterogeneity in cellular populations is crucial for disease treatment and population manipulation. In this study, we propose a Bayesian inference framework to elucidate sources of cell-to-cell variability in yeast signaling, providing deeper insights into the causes and consequences of heterogeneity.
PLOS COMPUTATIONAL BIOLOGY
(2022)
Article
Mathematical & Computational Biology
Ellen C. Caniglia, Eleanor J. Murray, Miguel A. Hernan, Zach Shahn
Summary: Methods for estimating optimal treatment strategies often assume unlimited resources, but practical constraints such as limited funds or medication access require considering competition for resources. While solving this problem for single exposure points is established, extending it to dynamic treatment strategies presents challenges.
STATISTICS IN MEDICINE
(2021)
Article
Environmental Sciences
Xiaole Zhang, Xiaoxiao Feng, Jie Tian, Yong Zhang, Zhiyu Li, Qiyuan Wang, Junji Cao, Jing Wang
Summary: Millions of premature mortalities are caused by PM2.5 air pollution globally each year. To address this issue, the researchers updated the emission inventories and harmonized the results from source-oriented and receptor models, finding a significant reduction in pollutant emissions in Beijing and successfully identifying temporal changes during the Spring Festival.
SCIENCE OF THE TOTAL ENVIRONMENT
(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
Engineering, Civil
Chao Dang, Marcos A. Valdebenito, Matthias G. R. Faes, Jingwen Song, Pengfei Wei, Michael Beer
Summary: This paper proposes a new method called 'Bayesian active learning line sampling' (BAL-LS), which derives the exact posterior variance of the failure probability to measure the epistemic uncertainty more accurately. In addition, the method proposes two essential components (learning function and stopping criterion) to facilitate Bayesian active learning based on the uncertainty representation of the failure probability. Compared with PBAL-LS, BAL-LS also has the advantage of automatically updating the important direction. Four numerical examples demonstrate the efficiency and accuracy of the proposed method in evaluating extremely small failure probabilities.
Article
Statistics & Probability
Weichang Yu, Howard D. Bondell
Summary: In this paper, a Bayesian likelihood-based dynamic treatment regime model is proposed, which incorporates regression specifications to interpret the relationships between covariates and stage-wise outcomes. A set of probabilistically-coherent properties for dynamic treatment regime processes are defined, and the theoretical advantages consequential to these properties are presented. Through a numerical study, it is demonstrated that the proposed method outperforms existing state-of-the-art methods.
JOURNAL OF THE ROYAL STATISTICAL SOCIETY SERIES B-STATISTICAL METHODOLOGY
(2023)
Article
Physics, Fluids & Plasmas
Kai Shimagaki, John P. Barton
Summary: This article proposes a framework for accurately estimating time-integrated quantities using Bezier interpolation and applies it to two dynamical inference problems. The results show that Bezier interpolation reduces estimation bias, especially for data sets with limited time resolution.
Article
Mathematics, Interdisciplinary Applications
Imke Botha, Robert Kohn, Christopher Drovandi
Summary: Parameter inference for SDEMEM models is challenging due to the lack of analytical solutions and intractable likelihood calculations. Particle MCMC methods offer exact inference possibilities, but naive implementations may be highly inefficient. This article introduces three extensions to improve inference efficiency by exploiting specific aspects of SDEMEM models and incorporating correlated pseudo-marginal methods. Comparisons on simulated and real data from a tumour xenography study demonstrate the effectiveness of these methods.
Article
Statistics & Probability
Shuxiao Chen, Bo Zhang
Summary: This study proposes a method to estimate dynamic treatment regimes (DTRs) with a time-varying instrumental variable (IV) in the presence of unmeasured confounding. The authors derive a novel Bellman equation to define a generic class of estimands, termed IV-optimal DTRs, and extend this framework to address the policy improvement problem. They demonstrate the superior performance of IV-optimal and IV-improved DTRs over DTRs that assume no unmeasured confounding.
JOURNAL OF THE ROYAL STATISTICAL SOCIETY SERIES B-STATISTICAL METHODOLOGY
(2023)
Article
Ecology
Johannes Oberpriller, David R. Cameron, Michael C. Dietze, Florian Hartig
Summary: Ecologists rely on complex computer simulations to forecast ecological systems, but uncertainties in model parameters and structure can lead to bias and underestimation. The article proposes a framework for robust inference and suggests solutions such as data rebalancing and bias corrections to improve model accuracy. Developing better methods for robust inference in complex computer simulations is crucial for generating reliable predictions of ecosystem responses.
Article
Mathematics
Francisco Javier Diez, Manuel Arias, Jorge Perez-Martin, Manuel Luque
Summary: OpenMarkov is an open-source software tool designed for probabilistic graphical models, primarily in medicine but also used in other fields and education in over 30 countries. This paper explains how OpenMarkov can be used as a pedagogical tool to teach the main concepts of Bayesian networks and influence diagrams, as well as various inference algorithms.
Article
Mathematical & Computational Biology
Liis Starkopf, Shahzleen Rajan, Theis Lange, Thomas Alexander Gerds
Summary: This paper investigates the estimation of the probability of a binary counterfactual outcome with a continuous covariate under monotonicity constraints. It focuses on studying out-of-hospital cardiac arrest patients and aims to estimate the counterfactual 30-day survival probability based on the ambulance response time and the presence of bystander cardiopulmonary resuscitation (CPR). The paper proposes a marginal structural model and B-splines to model the monotone relationship, and utilizes an auxiliary regression model for the observed 30-day survival probabilities to derive an estimating equation for the parameters of interest. The methods are demonstrated and compared with an unconstrained modeling approach using large-scale Danish registry data.
STATISTICS IN MEDICINE
(2023)
Article
Biology
Zeyu Bian, Erica E. M. Moodie, Susan M. Shortreed, Sahir Bhatnagar
Summary: Dynamic treatment regimes (DTRs) are a sequence of decision rules that recommend treatments for individual patients based on their information history. With increasing data complexity, it can be challenging to identify relevant prognostic factors in the treatment rule. Therefore, we propose a data-driven variable selection method to improve the estimation of decision rules.
Article
Mathematics, Interdisciplinary Applications
Erica E. M. Moodie, Janie Coulombe, Coraline Danieli, Christel Renoux, Susan M. Shortreed
Summary: This study discusses the estimation of individualized treatment rules for depression treatment and privacy protection, using distributed regression and dynamic weighted survival modeling to blur individual-level data and address the challenges of data privacy and small treatment effect heterogeneity.
LIFETIME DATA ANALYSIS
(2022)
Article
Biology
Armando Turchetta, Erica E. M. Moodie, David A. Stephens, Sylvie D. Lambert
Summary: This study provides a Bayesian approach for calculating sample size, allowing for more accurate and robust estimates that account for uncertainty in inputs through the "two priors" method. Compared to standard frequentist formulae, this methodology relies on fewer assumptions, incorporates pre-trial knowledge, and shifts the focus to the MDD.
Article
Mathematical & Computational Biology
Eric J. Rose, Erica E. M. Moodie, Susan M. Shortreed
Summary: Data-driven methods for personalizing treatment assignment have received attention from clinicians and researchers. Dynamic treatment regimes use decision rules to recommend individualized treatments. Observational studies are commonly used due to cost constraints, but can lead to biased estimation of treatment regimes. Sensitivity analyses, such as Monte Carlo sensitivity analysis, can assess the robustness of study conclusions. We propose a method for performing a Monte Carlo sensitivity analysis of unmeasured confounding in the estimation of dynamic treatment regimes, demonstrating its performance with simulations and an observational study on antidepressant medication.
BIOMETRICAL JOURNAL
(2023)
Editorial Material
Immunology
Erica E. M. Moodie
JOURNAL OF INFECTIOUS DISEASES
(2023)
Article
Statistics & Probability
Widemberg S. Nobre, Alexandra M. Schmidt, Erica E. M. Moodie, David A. Stephens
Summary: We propose a Bayesian procedure to estimate causal effects for multilevel observations in the presence of confounding. The motivation is to determine the causal impact of directly observed therapy on the successful treatment of Tuberculosis. We discuss the inclusion of latent local-level random effects in the propensity score model to reduce bias in the estimation of causal effects. A simulation study suggests that accounting for the multilevel nature of the data with latent structures in both the outcome and propensity score models has the potential to reduce bias in the estimation of causal effects.
JOURNAL OF THE ROYAL STATISTICAL SOCIETY SERIES C-APPLIED STATISTICS
(2023)
Article
Mathematical & Computational Biology
Erica E. M. Moodie, Zeyu Bian, Janie Coulombe, Yi Lian, Archer Y. Yang, Susan M. Shortreed
Summary: Despite increasing interest in individualized treatment rules, there is little attention given to binary outcomes. This study introduces a new computational approach to estimate a doubly robust regularized equation for binary outcomes in depression treatment. The approach demonstrates its double robustness and effectiveness in variable selection. The study is motivated by and applied to the analysis of treatment for unipolar depression using data from patients treated at Kaiser Permanente Washington.
Article
Mathematics, Interdisciplinary Applications
Xiao Li, Brent R. Logan, S. M. Ferdous Hossain, Erica E. M. Moodie
Summary: In order to provide the best possible care to each individual, physicians need to customize treatments based on their health state, especially for diseases like cancer that require additional treatments as they progress. This article introduces a Bayesian machine learning framework, using Bayesian additive regression trees (BART) to optimize dynamic treatment regimes (DTRs) for censored outcomes. The proposed approach is compared with Q-learning using simulation studies and a real data example.
LIFETIME DATA ANALYSIS
(2023)
Article
Mathematical & Computational Biology
Shuo Sun, Johanna G. Neslehova, Erica E. M. Moodie
Summary: This study proposes an approach to estimate quantile causal effects within a principal stratification framework, using a conditional copula approach to impute missing compliance data. The feasibility of this method is verified through a simulation study, and the approach is applied to investigate the impact of stay-at-home orders on COVID-19 transmission risk.
STATISTICS IN MEDICINE
(2023)
Article
Mathematical & Computational Biology
Armando Turchetta, Nicolas Savy, David A. A. Stephens, Erica E. M. Moodie, Marina B. B. Klein
Summary: Forecasting recruitments is crucial in the monitoring phase of multicenter studies. The Poisson-Gamma recruitment model is a popular technique based on the doubly stochastic Poisson process. However, the assumption of constant recruitment rates is often unrealistic in real studies. This paper presents a flexible generalization of the model, allowing varying enrollment rates over time using B-splines. The approach is shown to be suitable for a wide range of recruitment behaviors in simulations and is applied to estimate recruitment progression in a Canadian Co-infection Cohort.
STATISTICS IN MEDICINE
(2023)
Article
Statistics & Probability
Zeyu Bian, Erica E. M. Moodie, Susan M. Shortreed, Sylvie D. Lambert, Sahir Bhatnagar
Summary: An individualised treatment rule (ITR) is a decision rule that improves individuals' health outcomes by recommending treatments based on subject-specific information. It is important to select variables to improve the treatment rule and exclude irrelevant variables to increase efficiency and simplify the rule.
JOURNAL OF THE ROYAL STATISTICAL SOCIETY SERIES C-APPLIED STATISTICS
(2023)