Investigation and Construction of Statistical Designs for Multifactorial Experiments

-----------------------------------------------------------------------提案编号：DMS 9803596 PI：Samad Hedayat研究所：伊利诺伊大学芝加哥分校项目：多因素统计设计的调查和构建实验摘要：研究人员在科学和工业实验的设计和分析领域研究统计问题，其中实验结果受多种因素和噪声的影响。目标是确定最佳因素组合，同时最小化和测量噪声。研究工作主要集中在能够研究位置和分散效应的设计、高强度的复合正交表、作为核心子设计的g设计和强度t+的正交表的构造。任何严肃的科学研究的成功都取决于它的设计方式以及收集和分析的相关数据。噪音是最大的滋扰，而目标是将结果与影响研究的因素的水平联系起来。这在几乎所有的社会、工程、环境和医学研究中都是如此。完全消除噪音是昂贵的，而且确实是不明智的。事实上，在许多研究中，只要知道噪音的来源和强度就足够了。在后一种情况下，可以探索周围因素是如何影响结果的。研究人员探索了所有这些相互关联的主题，并确定了能够处理这些高度复杂的科学问题的新的具有成本和时间效益的改进统计设计。

----------------------------------------------------------------------- Proposal Number: DMS 9803596 PI: Samad Hedayat Institution: University of Illinois at Chicago Project: Investigation and Construction of Statistical Designs for Multifactorial Experiments Abstract: The investigator studies statistical problems in the area of design and analyses of scientific and industrial experiments in which the outcomes are influenced by multiple factors and noise. The goal will be to identify the best factor combinations while minimizing and measuring the noise. The research effort is mainly concentrated in the construction of designs capable of investigating location and dispersion effects, compound orthogonal arrays with high strength, g-designs as core sub-designs ,and orthogonal arrays of strength t plus. The success of any serious scientific study depends on the way it is designed and the related data being collected and analyzed. Noise is the biggest nuisance while the goal is to relate the outcomes to the levels of the factors influencing the study. This is true in almost all social, engineering, environmental, and medical studies. It is costly and indeed unwise to eliminate noise entirely. Indeed in many studies it will be enough if one knows the source and the intensity of noise. Under the latter condition it is possible to explore how the surrounding factors are impacting the outcomes. The investigator explores all these interconnected topics and identifies new cost- and time-effective and improved statistical designs capable of handling these highly complicated scientific issues.