Abstract
Generally, among many SQC methods, the multiple regression analysis is used most to solve problems because of its convenience and effectiveness. But it has various problems itself, and no well-practicable method to cope with these problems have been proposed. Therefore, it is seldom utilized. This paper focuses on the variation source analysis that is the starting point of applied multiple regression analysis. First, its importance and then the problems arising from the influence of multicollinearity are studied. Then, the BN method (the forward stepwise BEST "N" method) is proposed. Next, the value of "N" on BEST "N" for selecting variables most effectively is searched by simulation. Finally, the effectiveness of BN method is demonstrated by applying it to actual cases.
