美国Indiana University South Bend数学科学系
报告摘要：
The presentation includes two parts. The first part is a brief introduction of adaptive group sequential designs in clinical trials. The second part presents a specific adaptive design, namely, Bayesian group sequential designs, summarized as follows.
A clinical trial requires decisions about how an experiment will be conducted. The process is often dynamic; therefore, Bayesian formulation provides a powerful mechanism for running such a process.
The decision theoretical approach considers the consequences of each possible designing parameter and selects ones to maximize the expected utility. While utility setting and statistical decisions depend on the goals of the experiment, certain choices may be restricted by available resources and ethical considerations.
The Bayesian adaptive group sequential designs also have the flexibility of allowing interpretation of the results along frequentist lines, in addition to maximize the utility function. The frequentist error rates can be established so that the desired properties can be maintained for regulatory settings.
