Modified sequential bifurcation for simulation factor screening under skew-normal response model

Abstract

Sequential bifurcation (SB) has been widely used for simulation factor screening problems to identify important factors (i.e., factors that can significantly affect system response) with high efficiency and effectiveness under the 'sparse effects' assumption. However, when the response is asymmetric, previous SB studies based on the assumption of normally distributed response have shown limitations in practice. In this paper, we develop a skew-normal distributed response model and propose a parametric bootstrap testing procedure that is incorporated into every 'bifurcation' stage of SB for testing the significance of factors. Numerical experiments are provided to demonstrate the effectiveness and efficiency of the screening methods.