Title: When design of experiment meets computing

Abstract: The design of experiment (DOE) is to collect the data in the most informative way for the scientific questions of interest at a minimum cost. Experiments are needed in scientific discovery, clinical trials, business operations, etc. In this talk, I would present my recent works on some "non-traditional" experiments. The first one is to study how to best select a batch of units out of a large pool. The second one is the efficient computing of an important type of statistics using DOE ideas. The third one, if time permitted, is to approximate the integration of an expensive to evaluate black box function efficiently by the DOE idea.