Block, Sampson and Savits 9704400 This research focuses on the development of methodology and theory for some major problems facing the statistics and reliability communities. One major emphasis is on the topic of burn-in. On-going research involves determining optimal burn-in for classes of distributions and families of cost structures. Of particular interest is the family of distributions with bathtub shaped hazard functions. A second major emphasis of this research is in the development of methods to support the use of the modern nonparametric ideas of order restricted inference in data analysis involving multivariate observations. A great deal of theory has been developed for order restricted inference; however, the research focus here is to utilize some of these results in important estimation and testing problems, in part, supported by the application of some of the computational results obtained by the investigators. This research makes effective use of this elegant class of nonparametric procedures in modern data analysis. Two other areas of research are the further development of models and methodologies for data arising from accelerated degradation experiments, and the continued development of the modeling of random length multivariate data. These four research areas, while dealing with varied problems, utilize theory and methods from the allied areas of reliability, order restricted inference and multivariate analysis. Burn-in is a widely used engineering practice for increasing reliability by eliminating weak industrial products before they are released for sale or use. This is an important topic in many modern American industrial processes. Statistical tools are being developed to implement new production strategies which are useful in determining how to deal with items which might fail early in critical uses. The clinical development of cost-effective new bioengineered pharmaceuticals often requires identification of those patients who would b enefit substantially more from new biopharmaceuticals than from standard available treatments. Some of this research focuses on developing improved statistical techniques to better identify those patients who would significantly benefit from the new biotechnology. Other aspects of the research involve obtaining more effective statistical approaches to allow clinicians to predict the likelihood of specific patients having a successsful response to a new treatment.
