Article
Psychology, Multidisciplinary
Qianrao Fu, Mirjam Moerbeek, Herbert Hoijtink
Summary: This paper introduces the R package SSDbain, which calculates the sample size required for evaluating hypotheses using Bayesian ANOVA. Researchers can express their expectations on group means through equality and order constrained hypotheses. The package provides tools for easily planning the sample size in the social and behavioral sciences.
FRONTIERS IN PSYCHOLOGY
(2022)
Article
Computer Science, Interdisciplinary Applications
Yueqi Shen, Matthew A. Psioda, Joseph G. Ibrahim
Summary: BayesPPD is an R package that enables Bayesian power and type I error calculation and model fitting for generalized linear models (GLM) by incorporating historical data and using power prior and normalized power prior. It supports summary level data or subject level data with covariate information, and allows the use of multiple historical datasets for design.
Article
Computer Science, Interdisciplinary Applications
Saroja Kumar Singh, Sarat Kumar Acharya, Frederico R. B. Cruz, Roberto C. Quinino
Summary: This study investigates the method to determine the sample size for an M/D/1 queueing system under the Bayesian setup by observing the number of customer arrivals during the service time. The proposed approach shows efficiency and efficacy in applications such as production management and telecommunications networks.
MATHEMATICS AND COMPUTERS IN SIMULATION
(2021)
Article
Neurosciences
Pravesh Parekh, Gaurav Vivek Bhalerao, John P. John, G. Venkatasubramanian
Summary: This study investigates the minimum sample size required for achieving multisite harmonization and provides insights on regressing the effect of covariates and the impact of Mahalanobis distance on harmonization.
Article
Mathematical & Computational Biology
Susanna Gentile, Valeria Sambucini
Summary: Classical power analysis is commonly used in clinical trials for sample size determination. In this paper, we explore and compare a hybrid classical Bayesian and a fully Bayesian approach to incorporate prior information and provide flexibility in the design assumptions. We utilize exact methods to establish the rejection of the null hypothesis and propose a conservative criterion for sample size determination in the presence of discrete data. A Shiny web app in R is developed to compute the optimal sample size according to the proposed criteria and ensure reproducibility of the results.
BIOMETRICAL JOURNAL
(2023)
Article
Mathematical & Computational Biology
Rebecca. M. M. Turner, Michelle. N. N. Clements, Matteo Quartagno, Victoria Cornelius, Suzie Cro, Deborah Ford, Conor. D. D. Tweed, A. Sarah Walker, Ian. R. R. White
Summary: Bayesian analysis in non-inferiority trials allows for direct probability statements about treatment differences, without relying on arbitrary non-inferiority margins. Three Bayesian approaches for determining sample size in non-inferiority trials with binary outcomes are presented, along with their advantages and disadvantages. These approaches include a predictive power approach, an expected posterior probability approach, and a precision-based approach. Applying these methods to an antiretroviral therapy trial for HIV-infected children, the predictive power approach is most accessible, but sample sizes are larger than with frequentist calculations. The expected posterior probability approach leads to smaller sample sizes and is suitable for estimating posterior probability. The precision-based approach is useful when recruitment or cost limitations exist, but may be difficult to decide on sample size alone using this approach.
STATISTICS IN MEDICINE
(2023)
Article
Biology
Haiyan Zheng, Thomas Jaki, James M. S. Wason
Summary: This paper develops Bayesian sample size formulae that can incorporate preexperimental information from multiple sources to support both the design and analysis of experiments comparing two groups. The methodology can be applied to various scenarios and provides exact solutions or a search procedure for sample size determination.
Article
Mathematical & Computational Biology
Haiyan Zheng, Michael J. Grayling, Pavel Mozgunov, Thomas Jaki, James M. S. Wason
Summary: Basket trials are increasingly used for evaluating new treatments in different patient subgroups. This paper proposes a Bayesian approach to determine sample size in basket trials, allowing information borrowing between similar subsets. The proposed approach yields comparable sample sizes for circumstances of no borrowing, and significantly reduces sample size when borrowing is enabled between commensurate subtrials. Examples and simulation studies demonstrate the feasibility and effectiveness of the proposed methodology.
Article
Psychology, Mathematical
Qianrao Fu, Herbert Hoijtink, Mirjam Moerbeek
Summary: This paper introduces the usage of the R package SSDbain to determine sample size for comparing two independent means, along with the implementation of Bayesian t-test and Welch's test. Psychological researchers can easily calculate the required sample size using the methods and tools provided in this paper.
BEHAVIOR RESEARCH METHODS
(2021)
Article
Computer Science, Software Engineering
Jong-Hee Chung, Yong-Bin Lim, Donghoh Kim
Summary: Determining the appropriate sample size in a balanced design of experiments is crucial for statistical inference. The R package BDEsize can compute the optimal sample size to detect a certain standardized effect size with a certain power level, and its diagnostic graphs provide information on the relationships between sample size, power, and detectable effect size, helping researchers identify effects sensitive to sample size and adjust accordingly.
Article
Multidisciplinary Sciences
Ben O'Neill
Summary: This article analyzes standard confidence intervals for the mean of a finite population and proposes a new method for sample size determination that takes into account the uncertainty of the parameter. By using a preliminary sample of data, this method increases the accuracy of inference.
Article
Health Care Sciences & Services
Duncan T. Wilson, Richard Hooper, Julia Brown, Amanda J. Farrin, Rebecca E. A. Walwyn
Summary: The method proposes a general framework for solving simulation-based sample size determination problems with multiple design parameters and conflicting criteria to be minimized. It utilizes a global optimization algorithm and a non-parametric regression model to approximate the true underlying power function, making it flexible and applicable to a wide range of simulation-based power estimation problems.
STATISTICAL METHODS IN MEDICAL RESEARCH
(2021)
Article
Psychology, Multidisciplinary
Samantha F. Anderson, Ken Kelley
Summary: Replication is central to scientific progress, and new approaches are needed to assess the success of replication studies beyond traditional dichotomous outcomes. Sample size planning is crucial for replication studies and should align with the specific goals and analysis methods.
PSYCHOLOGICAL METHODS
(2022)
Article
Statistics & Probability
Sajid Ali, Mariyam Waheed, Ismail Shah, Syed Muhammad Muslim Raza
Summary: Sample size determination is an important topic in statistics research. Bayesian methods generally require smaller sample sizes compared to classical techniques, particularly when using the average length criterion. This study aims to utilize Bayesian techniques for determining the sample size for the coefficient of variation of a normal distribution and compare the results with frequentist approaches. The findings show that Bayesian methods require smaller sample sizes to achieve the same efficiency as frequentist methods.
JOURNAL OF APPLIED STATISTICS
(2023)
Article
Medicine, Research & Experimental
Jingxian Lan, Amy C. Plint, Stuart R. Dalziel, Terry P. Klassen, Martin Offringa, Anna Heath
Summary: This study used a remote, real-time expert elicitation method to construct prior distributions for international randomized controlled trials. Experts from various countries were recruited to determine the sample size of a trial for bronchiolitis in infants. The results showed that remote expert elicitation is a feasible and practical approach.
Review
Mathematical & Computational Biology
Leonhard Held, Robert Matthews, Manuela Ott, Samuel Pawel
Summary: It is widely believed that the current inferential toolkit used in scientific research is inadequate, and there is no consensus on alternative methods. A new Reverse-Bayes approach has been proposed to address the longstanding issues in Bayesian analysis, potentially providing solutions to inferential challenges and making Bayesian methods more accessible and attractive for evidence assessment.
RESEARCH SYNTHESIS METHODS
(2022)
Article
Cardiac & Cardiovascular Systems
Ladina Erhart, Beat A. Kaufmann, Baris Gencer, Philipp K. Haager, Hajo Mulller, Richard Kobza, Leonhard Held, Simon F. Staempfli
Summary: This study provides evidence that impairment in renal function is associated with an increased risk of death and heart transplantation in LVNC patients, suggesting that kidney function assessment should be standard in risk assessment of LVNC patients.
CARDIOLOGY JOURNAL
(2023)
Editorial Material
Biochemistry & Molecular Biology
Ulrich Mansmann, Clara Locher, Fabian Prasser, Tracey Weissgerber, Ulrich Sax, Martin Posch, Evelyne Decullier, Ioana A. A. Cristea, Thomas P. A. Debray, Leonhard Held, David Moher, John P. A. Ioannidis, Joseph S. S. Ross, Christian Ohmann, Florian Naudet
Summary: Data sharing improves the value of medical research and promotes trust in clinical trials, but more biomedical researchers need training in approaches such as meta-research, data science, and ethical, legal, and social issues.
Article
Mathematical & Computational Biology
Samuel Pawel, Lucas Kook, Kelly Reeve
Summary: Comparative simulation studies are crucial for evaluating statistical methods, but their validity can be compromised by questionable research practices. We propose concrete suggestions for enhancing the methodological quality of simulation studies, including preregistering protocols, incentivizing neutral studies, and promoting code and data sharing.
BIOMETRICAL JOURNAL
(2023)
Article
Infectious Diseases
Aline Wolfensberger, Lauren Clack, Stefanie von Felten, Mirjam Faes Hesse, Dirk Saleschus, Marie-Theres Meier, Katharina Kusejko, Roger Kouyos, Leonhard Held, Hugo Sax
Summary: This study aimed to test a prevention intervention for non-ventilator-associated hospital-acquired pneumonia and a multifaceted implementation strategy. The results showed that implementing the prevention intervention significantly reduced the incidence rate of nvHAP.
LANCET INFECTIOUS DISEASES
(2023)
Article
Pharmacology & Pharmacy
Manja Deforth, Charlotte Micheloud, Kit C. Roes, Leonhard Held
Summary: Conditional or accelerated approval of drugs allows earlier access to promising new treatments that address unmet medical needs. Traditional methods like Fisher's criterion and Stouffer's method can be used to support the design and analysis of post-market trials, but the harmonic mean chi(2)-test always requires a post-market clinical trial and may require a smaller sample size if the p-value from the pre-market clinical trial is << 0.025.
PHARMACEUTICAL STATISTICS
(2023)
Article
Statistics & Probability
Charlotte Micheloud, Fadoua Balabdaoui, Leonhard Held
Summary: We propose a statistical framework for replicability based on the sceptical p-value, which is a quantitative measure of replication success. A recalibration is suggested to achieve exact Type-I error control in the case of null effect in both studies, with additional bounds on the partial and conditional Type-I error rate. This approach avoids the need for double dichotomization and has higher power to detect existing effects across both studies. It can also be used for power calculations and requires a smaller replication sample size compared to the two-trials rule for convincing original studies. The performance of the proposed methodology is illustrated using data from the Experimental Economics Replication Project.
STATISTICA NEERLANDICA
(2023)
Article
Statistics & Probability
Samuel Pawel, Frederik Aust, Leonhard Held, Eric-Jan Wagenmakers
Summary: Power priors are used to incorporate historical data into Bayesian analyses, but a new theoretical result shows that when the current data perfectly mirror the historical data and both sample sizes become large, the marginal posterior distribution of alpha does not converge to a point mass at alpha=1, but approaches the prior distribution instead. This implies that complete pooling of historical and current data is impossible when using a power prior with a beta prior for alpha.
Article
Statistics & Probability
Samuel Pawel, Alexander Ly, Eric-Jan Wagenmakers
Summary: We propose a new and user-friendly method for calibrating error-rate based confidence intervals to evidence-based support intervals. Support intervals are obtained by inverting Bayes factors based on parameter estimate and standard error. We present different types of support intervals that allow analysts to encode external knowledge. We also demonstrate how to determine sample size for future studies based on support. The importance of the method is illustrated through an application to clinical trial data.
AMERICAN STATISTICIAN
(2023)
Article
Pharmacology & Pharmacy
Annette Kopp-Schneider, Manuel Wiesenfarth, Leonhard Held, Silvia Calderazzo
Summary: Borrowing information from historical or external data for inference in current trials is a growing field in precision medicine. This study proposes a procedure to investigate and report the operating characteristics of borrowing methods. The findings suggest that borrowing external data may not improve the power of certain trials.
PHARMACEUTICAL STATISTICS
(2023)
Article
Multidisciplinary Sciences
D. van Ravenzwaaij, M. Bakker, R. Heesen, F. Romero, N. van Dongen, S. Cruwell, S. M. Field, L. Held, M. R. Munafo, M. M. Pittelkow, L. Tiokhin, V. A. Traag, O. R. van den Akker, A. E. van't Veer, E. J. Wagenmakers
Summary: Theoretical arguments and empirical investigations suggest that a significant number of published findings cannot be replicated and are likely to be false. This paper presents a comprehensive perspective on scientific error, focusing on reform history and future opportunities. It discusses institutional reform, methodological reform, statistical reform, and publishing reform, providing potential errors through the narrative of a fictional researcher. The resulting agenda aims to create a research culture with fewer errors and a scientific publication landscape with fewer false findings.
ROYAL SOCIETY OPEN SCIENCE
(2023)