New Sliced Space-Filling Designs Construction and Theory
Technical Report,29 May 2015,28 Feb 2016
University of Wisconsin - Madison MADISON United States
Pagination or Media Count:
The purpose of this project is to improve sliced Latin hypercube designs which were previously developed by the PI and were successfully used by various DoD divisions including US Army. The project has successfully developed some new classes of statistical designs for computer simulation, uncertainty quantification, sensitivity analysis and stochastic optimization. These include bi-directional sliced Latin hypercube designs and sliced Latin hypercube designs with controlled correlations. The project also derived new statistical theories to study sampling properties of the proposed designs and their use in integration and stochastic optimization. Specific applications of the designs include running simulation codes with both quantitative and qualitative factors, running simulation experiments in batches and ensemble evaluation of multiple computer models. In addition, some experimental design ideas have motivated the development of new computational algorithms for accelerating large-scale statistical computing.
- Numerical Mathematics