Response adaptive designs with multiple objectives

点击翻译按钮获取中文摘要

I have spent much of my research career investigating optimal designs that are practically applicable in the real world. My research interests continue to be in constructing optimal designs, while dealing with various practical challenges. Building on my prior work, I plan to further work on building response adaptive designs for binary responses. The project concerns the role of statistical design, modelling, and analysis in quality experiments with repeated measures data. The primary purpose of this project is to continue the development of new statistical tools and methods that address real issues of concern in empirical scientific research. ******Traditionally, optimal design literature has focused on achieving a single objective, namely minimizing the mean square error of the estimated effects. However, considering significant benefits attainable by having multiple objectives satisfied, I find that there is really no going back to the tradition. More research is needed to build optimal designs for small samples adaptively on responses, placing efforts to reassess the optimality and allocating more patients to beneficial treatment(s). As well, response adaptive designs with continuous responses can be benefited by accommodating possible solutions on issues such as measurement errors and missing data methods. ******In our recent work “Multiple-objective response adaptive repeated measure designs in clinical trials for binary responses”, we discussed unique properties in binary designs that arise due to the error matrix being dependent on the parameters. No doubt, this dependency can play a significant role in the resulting optimal design. Misspecification of a covariance matrix may impact the design efficiency, and it will be a focus in this part of my research. Because these experimentations are usually expensive, emphasis is given to techniques and strategies which minimize experimental effort, and maximize information gained from analysis of the results. This research will construct designs that are robust to various experimental conditions with better understanding of possible conflicts between statistical and practical optimality. The foundations of practical optimal designs will be further developed. A variety of other statistical techniques to deal with real issues with repeated measures data will be explored and developed.**