Article
Mathematical & Computational Biology
Cyrus Mehta, Apurva Bhingare, Lingyun Liu, Pralay Senchaudhuri
Summary: This paper develops optimal decision rules for sample size re-estimation in two-stage adaptive group sequential clinical trials. The initial sample size specification of such trials is usually adequate to detect a realistic treatment effect with good power but insufficient to detect the smallest clinically meaningful treatment effect. It is difficult for sponsors to make an up-front commitment to power a study adequately to detect the smallest effect. If the interim data enter a promising zone, it is easier to justify increasing the sample size.
STATISTICS IN MEDICINE
(2022)
Review
Respiratory System
Elizabeth G. Ryan, Dominique-Laurent Couturier, Stephane Heritier
Summary: The use of Bayesian adaptive designs in clinical trials has increased recently, especially during the COVID-19 pandemic. Bayesian adaptive designs offer a flexible and efficient framework and may provide more useful and natural results for clinicians compared to traditional approaches.
Article
Statistics & Probability
Ming-Hung Kao, Hazar Khogeer
Summary: This study focuses on optimal designs for multivariate regression of responses of mixed variable types on quantitative and qualitative factors. New complete class results are derived using Loewner ordering and relevant Chebyshev systems, which identify a small class of designs where locally optimal designs can be found for a group of models and optimality criteria. The complete class results facilitate the search for optimal designs using general-purpose optimization techniques, and extensions of previous results for characterizing optimal designs are also provided.
JOURNAL OF MULTIVARIATE ANALYSIS
(2021)
Article
Automation & Control Systems
Mingduo Lin, Bo Zhao, Derong Liu
Summary: A model-free optimal tracking controller for discrete-time nonlinear systems is designed using policy gradient adaptive critic designs and experience replay, aiming to improve control performance.
IEEE TRANSACTIONS ON SYSTEMS MAN CYBERNETICS-SYSTEMS
(2022)
Article
Mathematics, Applied
Duc-Lam Duong, Tapio Helin, Jose Rodrigo Rojo-Garcia
Summary: We study the stability properties of the expected utility function in Bayesian optimal experimental design. We provide a framework for this problem in a non-parametric setting and prove the convergence rate of the expected utility with respect to a likelihood perturbation. The assumptions set out for the general case are satisfied in the specific case of non-linear Bayesian inverse problems with Gaussian likelihood, and the stability of the expected utility with respect to perturbations to the observation map is regained. Numerical simulations are used to demonstrate the theoretical convergence rates in three different examples.
Article
Management
Arielle Anderer, Hamsa Bastani, John Silberholz
Summary: This paper proposes a method of combining data from both surrogate and true outcomes to improve decision making within late-phase clinical trials, and introduces a Bayesian adaptive clinical trial design. Compared to traditional clinical trial designs, this approach is estimated to decrease trial costs by 16% while maintaining the same error rates.
MANAGEMENT SCIENCE
(2022)
Article
Mathematical & Computational Biology
Maximilian Pilz, Kevin Kunzmann, Carolin Herrmann, Geraldine Rauch, Meinhard Kieser
Summary: The study demonstrates that optimal adaptive designs can bring significant benefits in clinical trial planning. By customizing the underlying optimization problem, optimal designs can be tailored to specific operational requirements for an almost negligible performance loss compared to conventional designs.
STATISTICS IN MEDICINE
(2021)
Article
Statistics & Probability
Jonathan Stallrich, Katherine Allen-Moyer, Bradley Jones
Summary: Traditionally, optimal screening designs for arbitrary run sizes are generated using the D-criterion with fixed factor settings at +/-1, but this article identifies cases where D-optimal designs have undesirable estimation variance properties. It argues that A-optimal designs generally minimize variances closer to their minimum value. New insights about the criteria are gained through studying their coordinate-exchange formulas. The study confirms the existence of D-optimal designs with only +/-1 settings for main effect and interaction models in both blocked and unblocked experiments. Arbitrary manipulation of a coordinate between [-1,1] can result in infinitely many D-optimal designs with different variance properties.
Article
Automation & Control Systems
Xiong Yang, Qinglai Wei
Summary: This article introduces an optimal event-driven control scheme for a continuous stirred tank reactor (CSTR) system, utilizing discounted cost and adaptive critic designs to solve the optimal event-driven control problem. Stability analysis of signals in the closed-loop system is conducted using the Lyapunov approach.
IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS
(2021)
Article
Multidisciplinary Sciences
Keno Juechems, Jan Balaguer, Bernhard Spitzer, Christopher Summerfield
Summary: When making economic choices, humans may distort their internal representation of the value and probability of a prospect, but under the assumption of finite computational precision, these distortions may be approximately optimal, helping to maximize reward and minimize uncertainty. Two empirical studies show that humans can adapt optimally to factors that change the reward-maximizing form of distortion, providing an answer to the question of why humans make irrational economic choices.
PROCEEDINGS OF THE NATIONAL ACADEMY OF SCIENCES OF THE UNITED STATES OF AMERICA
(2021)
Review
Pharmacology & Pharmacy
Xin Chen, Ruyue He, Xinyi Chen, Liyun Jiang, Fei Wang
Summary: Small sample sizes in early-phase clinical trials may not establish clear toxicity and efficacy profiles for dose-schedule regimens. Bayesian adaptive designs offer an effective way to evaluate multiple doses and schedules in a single trial, but their real-world implementation examples are limited. This paper reviews critical factors and innovative Bayesian adaptive designs for dose-schedule optimization, and discusses unresolved issues and future research opportunities.
FRONTIERS IN PHARMACOLOGY
(2023)
Article
Mathematics
Necla Gunduz, Bernard Torsney
Summary: This paper focuses on proving a conjecture made by Ford, Torsney, and Wu regarding the existence of a class of D-optimal designs for binary and weighted linear regression models. The conjecture states that there exists a two-level factorial design that is D-optimal for these models. To prove this, we use an intuitive approach that explores various link functions in the generalized linear model context. Our results establish the existence of D-optimal designs for binary and weighted linear regression models with one design variable, which have important implications for the efficient design of experiments.
Article
Oncology
Emily C. Zabor, Alexander M. Kaizer, Nathan A. Pennell, Brian P. Hobbs
Summary: This study proposes three randomized designs for early phase biomarker-guided oncology clinical trials, using optimal efficiency predictive probability method to monitor multiple biomarker subpopulations for futility. The simulation study demonstrates that potentially smaller phase II trials can be designed efficiently using randomization and futility stopping to obtain more information before phase III studies.
FRONTIERS IN ONCOLOGY
(2022)
Article
Operations Research & Management Science
Vu Thi Huong, Jen-Chih Yao, Nguyen Dong Yen
Summary: We investigate a type of finite horizon optimal economic growth problems with nonlinear utility functions and linear production functions. Through the application of the maximum principle in optimal control theory and leveraging the special structure of the problems, we are able to explicitly describe the unique solution along with the input parameters. Economic interpretations of the results obtained and an open issue concerning the scenario where the total factor productivity falls within a bounded open interval specified by the growth rate of labor force, the real interest rate, and the exponent of the utility function are also discussed.
JOURNAL OF OPTIMIZATION THEORY AND APPLICATIONS
(2021)
Article
Ecology
Clinton B. Leach, Perry J. Williams, Joseph M. Eisaguirre, Jamie N. Womble, Michael R. Bower, Mevin B. Hooten
Summary: Optimal design procedures, when combined with Bayesian hierarchical models and recursive Bayesian computation, offer efficient tools for ecological learning and inference while reducing computational burdens. Examples using simulated data and real-world cases demonstrate the practical application of this method and emphasize the importance of computational gains for monitoring and science integration.