9504041 Hardwick This proposal concerns the use of parallel computers to help solve computationally intensive problems in statistical design and analysis. The designs considered are sequential, meaning that a design parameters can be altered during the experiment in response to the observations already gathered. Such designs can offer significant cost benefits, but are insufficiently utilized due to the complex analyses they require and the fact they are complicated to conduct. This research attacks these difficulties by considering procedures that are significantly easier to conduct, yet nearly as optimal. The computations are still a major hurdle, especially because the problem one needs to solve require combining sequential designs and computationally intensive bootstrapping.
