Design and analysis of repeated measures data

我的研究项目一直在重复测量数据的设计与分析的一般领域，并考虑了缺失数据分析、固定最优设计构建、响应自适应设计等问题。为了扩大范围，我最近考虑了一个更一般的模型来构建最优设计，因为最优设计强烈依赖于模型。我还考虑适应数据中固有的异质性。在本提案中，我的研究计划的主要目标之一是将单目标设计扩展到多目标设计，以构建响应自适应设计。我计划开发一种新的自适应分配规则，可以准确和最优地估计治疗效果对比。这项工作的应用被发现在许多领域，如工程、农业、生物科学以及生物医学科学。具体来说，我将在一般模型下构建重复测量设计，并记录它们对传统优化设计的影响。一般模型将把结转效应分为两种不同的类型，一种是由于在前一时期实施了相同的治疗措施，另一种是由于在前一时期实施了其他竞争性治疗措施。迄今为止构建的大多数最佳设计都假设这两种类型是相等的。此外，它们是基于固定的样本量。我将应用拉格朗日乘数法，并给出构建两个处理p周期重复测量设计的建议。将总结具有较少重复周期的实用设计。

My research program has been in the general area of design and analysis of repeated measures data, and has considered such issues as missing data analysis, fixed optimal design construction, and response adaptive designs. To broaden the scope, I have recently considered a more general model to construct optimal designs, as optimal designs are strongly model dependent. I also considered accommodating the heterogeneity inherent in the data.In this proposal, one of the main goals of my research program is to extend the single-objective designs to multiple-objective designs to build response-adaptive designs. I plan to develop a new adaptive allocation rule that can provide estimates of the treatment effect contrasts accurately and optimally. Application of this work is found in a number of fields such as engineering, agriculture, biosciences, as well as biomedical sciences.Specifically, I will construct repeated measurement designs under general models and document their impact on the traditional optimal designs. The general model will separate the carryover effects into two different types, one that results from the administration of the same treatment in the previous period and the other that results from the administration of the other competing treatment in the previous period. Most optimal designs constructed to date assume that these two types are equal. Further, they are based on fixed sample sizes. I will apply the Lagrange multiplier method and give recommendations on constructing two treatment p-period repeated measurement designs. Practically useful designs with fewer repeated periods will be summarized.