Grant: Discovery Grants Program - Individual

For the next few years, the main focus of my research will be on developing statistical techniques for constructing and analyzing designs (sets of input settings) for industrial experiments and experiments involved in investigating complex physical phenomena. In many branches of physical science (e.g., oceanography, climatology), when complex physical phenomena are very expensive or time-consuming to observe, observations from their corresponding computer models are used in addition for a better understanding of the underlying processes. An interesting research problem is to design experiments for the combined process, which is desired due to limited resources and time constraints. I also plan to use the growing computing power (e.g., parallel programming tools) to find designs that are optimal for experiments with pre-determined goals. This research will be useful in many areas of physical science when the true physical phenomena are either very expensive or infeasible to observe. For example, atmospheric scientists may wish to identify target levels of man-made pollutants that result in global temperature change of no more than 2 degrees celsius. Similarly, limitations on resources in industrial settings generates the demand for constructing useful experimental designs. In addition, it is often desirable to impose restrictions on the way experimental runs are performed, for example, in ndustrial experiments with multiple stages. It turns out that good designs for such experiments is not straightforward to find, however can be constructed using some finite projective geometric tools. I intend to develop both the theoretical results that confirms the existence and algorithms for constructing efficient multi-stage designs that are useful for practitioners.