Abstract
Screening experiments are often conducted before optimization in order to reduce computation resources by identifying the important factors of the problem. In the literatures, factor screening and simulation optimization approaches mostly adopted expectation as performance measures. The methodologies that are focused on other alternatives, however, are difficult to develop due to a lack of nice statistical properties as expectation. Quantile is an important alternative to the expectation for spatial data and moreover, it enables risk control. In this study, we propose a novel approach called STRONG-Q that integrates efficient quantile-based factor screening methods into the framework of STRONG, which is a newly-developed Response-Surface-based framework, for large-scale quantile-based simulation optimization problems. The quantile-based factor screening method can effectively control the Type I error and enables the large-scale quantile-based simulation optimization problems to be solved efficiently when it is integrated into STRONG.