다변량 분산분석을 이용한 다중 성능 시스템의 인자 스크리닝 기법

Abstract

In design process, using all the design variables can increase degrees of freedom for design and help to improve performance of the design. However, considering all the design variables requires a lot of time and high cost. Thus, it is necessary to employ a screening method that selects the design variables which are significant for the performance. Analysis of variance (ANOVA) is one of the most commonly used screening methods. However, ANOVA is performed for each performance independently, thus it might increase the type 1 error of the results. Multivariate analysis of variance (MANOVA) analyzes multiple performances at a time so it prevents error from inflating by multiple ANOVAs. Therefore, MANOVA which is for testing the significance of design variables gives more reliable results than previous multiple ANOVAs. This paper proposes MANOVA as a design variable screening method and verifies the usefulness of MANOVA by comparing results of multiple ANOVAs and MANOVA through the example.